find-skills
vercel-labs/skillsHelps the agent find and install agent skills from the open ecosystem with the skills CLI.
Agent workflow skills for Claude Code.
Most installed first. Copy the install command into a terminal; Claude Code loads the skill when a task needs it.
grill-with-docs
mattpocock/skillsInterviews you relentlessly about a plan or design while recording decisions in glossary and ADR files.
setup-matt-pocock-skills
mattpocock/skillsScaffolds per-repo config for the engineering skills: issue tracker, triage labels, and domain doc layout.
handoff
mattpocock/skillsCompacts the current conversation into a handoff document saved to a temp folder for another agent.
grilling
mattpocock/skillsInterviews the user in rounds about a plan, giving a recommended answer to each question until aligned.
triage
mattpocock/skillsMoves issues and external PRs through triage roles, verifies them, and writes agent-ready briefs.
implement
mattpocock/skillsImplements a spec or tickets, runs tdd and code review, then commits to the current branch.
ask-matt
mattpocock/skillsRoutes you to the right skill or workflow for your situation, from idea to shipped code.
wayfinder
mattpocock/skillsMaps a too-large effort as decision tickets on the tracker and resolves them until the route is clear.
to-tickets
mattpocock/skillsSplits a plan into vertical tracer-bullet tickets with blocking edges and publishes them to the tracker.
caveman
JuliusBrussee/cavemanMakes the agent reply in terse, caveman-style prose to cut output tokens while keeping technical facts.
skill-creator
anthropics/skillsGuides the agent through creating, testing, benchmarking and optimizing other skills with evals.
caveman-compress
JuliusBrussee/cavemanCompresses a memory file into caveman-style text to cut input tokens, keeping a backup of the original.
brainstorming
obra/superpowersMakes the agent clarify what the user really wants and get a design approved before building anything.
caveman-help
JuliusBrussee/cavemanShows a quick reference card for the caveman skills, modes and commands without changing any state.
cavecrew
JuliusBrussee/cavemanGuides the agent to hand code search, small edits and diff review to subagents with compressed output.
caveman-stats
JuliusBrussee/cavemanPrints the recorded token usage of the current Claude Code session without estimating savings.
orchestration
stablyai/orcaCoordinates supervised Orca workers and task DAGs, or hands work off to another agent via orca-cli.
implement-spec
mattpocock/skillsImplements a spec's ticket graph with parallel subagents on one integration branch.
retro
mattpocock/skillsReviews a coding session and suggests environment fixes such as navigation pointers and automated checks.
writing-plans
obra/superpowersMakes the agent turn a spec into a detailed, bite-sized implementation plan before touching code.
using-superpowers
obra/superpowersMakes the agent check for and invoke relevant skills before any response or action.
verification-before-completion
obra/superpowersMakes the agent run fresh verification commands and read their output before claiming work is done.
executing-plans
obra/superpowersHas the agent execute an implementation plan task by task in the current session, with one final whole-branch review.
subagent-driven-development
obra/superpowersHas the agent run an implementation plan by dispatching a fresh implementer per task, with reviews after each and at the end.
dispatching-parallel-agents
obra/superpowersMakes the agent split independent problems across parallel subagents with isolated, hand-built context.
writing-skills
obra/superpowersMakes the agent write, test and refine skills using a test-driven, RED-GREEN-REFACTOR style process.
caveman-explore
JuliusBrussee/cavemanRead-only repo explorer that locates relevant code and returns path and line citations only.
ponytail
DietrichGebert/ponytailMakes the agent solve each coding task with the smallest complete change and a brief, clear reply.
caveman-learn
JuliusBrussee/cavemanApplies caveman learn findings as consent-gated edits that cut agent token cost, measured per change.
herdr
herdrdev/herdrInspects and controls Herdr panes, tabs, and workspaces, but only when the user explicitly asks for Herdr.
spec-driven-development
addyosmani/agent-skillsMakes the agent write a structured spec with objectives and scope before coding, reviewed in gated phases.
planning-and-task-breakdown
addyosmani/agent-skillsSplits a spec into small ordered tasks with acceptance criteria and a dependency map.
context-engineering
addyosmani/agent-skillsCurates the context an agent sees: rules files, specs, source, and errors, loaded at the right time.
diagnosing-superpowers
obra/superpowersReads current or past session transcripts and builds an evidence-cited report on what went wrong in a superpowers session.
karpathy-guidelines
multica-ai/andrej-karpathy-skillsGuides the coding agent to state assumptions, keep changes minimal and define verifiable success criteria.
discernment-nudge
anthropics/skillsAfter substantive answers, appends two or three follow-up questions to help the user check facts and assumptions.
caveman
DietrichGebert/ponytailSwitches replies to terse caveman style, cutting about 75% of tokens while keeping technical accuracy.
ponytail-gain
DietrichGebert/ponytailShows ponytail's published benchmark savings in code, cost, and speed, without changing any state.
ponytail-help
DietrichGebert/ponytailShows a quick reference of ponytail levels, skills, and commands without changing any state.
autoplan
garrytan/gstackRuns CEO, design, engineering and DX reviews of a plan in sequence with automatic decisions.
captain
garrytan/gstackCoordinates a crew of coding agents from one controller that merges work in trains and relays approvals.
careful
garrytan/gstackWarns before running destructive shell commands such as recursive deletes or dropping database tables.
context-restore
garrytan/gstackReloads context saved by context-save so the agent can resume earlier work where it stopped.
context-save
garrytan/gstackSaves the current working state, decisions and next steps so a later session can pick them up.
gstack
garrytan/gstackRoutes a request to the right gstack skill in the suite.
doubt-driven-development
addyosmani/agent-skillsMakes the agent cross-examine non-trivial decisions with a fresh-context adversarial review before proceeding.
interview-me
addyosmani/agent-skillsInterviews the user one question at a time to pin down their real intent before any plan or code.
company-creator
paperclipai/paperclipScaffolds an agent company package that conforms to the agentcompanies/v1 spec, from scratch or from a repo.
create-case
paperclipai/paperclipCreates and kicks off a Paperclip case from a user request, with workflow instructions and an initial agent.
create-paperclip-bundled-skill
paperclipai/paperclipTurns a tweet, task, or idea into a reviewed skill in the Paperclip skills catalog, reusing prior art first.
create-verification-skill
paperclipai/paperclipGenerates a project-local skill that drives the app like a user and captures evidence it works.
paperclip
paperclipai/paperclipRuns Paperclip heartbeat tasks: reads task context, delivers documents, reports blockers and delegates work.
paperclip-board
paperclipai/paperclipManages a Paperclip AI-agent company via chat: onboarding, agents, approvals, task monitoring and costs.
paperclip-converting-plans-to-tasks
paperclipai/paperclipDecides whether plan steps need separate tasks and sets real blocking dependencies between them.
paperclip-create-agent
paperclipai/paperclipCreates Paperclip agents under hiring permission rules, comparing adapter configs and drafting a hire request.
para-memory-files
paperclipai/paperclipStores and retrieves durable facts, daily notes and entity records across sessions in a PARA file layout.
Agent Cost Report
thedotmack/claude-memBuilds a Claude Code cost report from local transcripts, estimated at list prices, as HTML and JSON.
build
thedotmack/claude-memChains make-plan, do and babysit skills to carry a coding task through without manual step-by-step prompts.
mem-search
thedotmack/claude-memSearches past claude-mem memory in layers, from index to full details, before reading raw tool logs.
dag-library
code-yeongyu/oh-my-openagentSaves a multi-agent DAG definition under a name so it can be re-run later without pasting it again.
hyperplan
code-yeongyu/oh-my-openagentRuns five adversarial agents to critique a plan, then turns the surviving insights into an executable plan.
i-have-adhd
ayghri/i-have-adhdShapes replies for an ADHD reader: next action first, numbered steps, restated state, and visible progress.
arbor
K-Dense-AI/scientific-agent-skillsRuns hypothesis-tree optimization loops that track experiments and evidence across repeated evaluator trials.
claude-md
luongnv89/claude-howtoCreates, updates, or audits CLAUDE.md files so AI agents get concise, accurate project onboarding.
self-assessment
luongnv89/claude-howtoRuns a multi-category quiz on Claude Code skills and returns per-topic scores and a personalized learning path.
claude-automation-recommender
anthropics/claude-plugins-officialReads a codebase and recommends hooks, subagents, skills, plugins and MCP servers to add, read-only.
claude-md-improver
anthropics/claude-plugins-officialAudits CLAUDE.md files against quality templates, reports gaps, then makes targeted updates.
writing-hookify-rules
anthropics/claude-plugins-officialExplains how to write hookify rule files that warn about or block agent actions by pattern.
parallel-feature-development
wshobson/agentsGuides splitting a feature into parallel work streams with file ownership and interface contracts for multiple agents.
task-coordination-strategies
wshobson/agentsGuides breaking complex tasks into parallel units with dependency graphs, clear task descriptions, and balanced workload.
team-communication-protocols
wshobson/agentsSets message-type rules, plan approval, and shutdown steps for communication between agent teammates.
team-composition-patterns
wshobson/agentsGuides choosing agent team size, preset team setups, and subagent types for a multi-agent task.
book-to-skill
virgiliojr94/book-to-skillTurns a book or document into a structured agent skill with frameworks, principles, techniques and anti-patterns.
multiagent-memory-setup
topoteretes/cogneeImports Claude Code and Codex session history into one cognee dataset and answers questions across both.
context-mode
mksglu/context-modeRoutes large command and tool output through a sandbox so only the needed results enter context.
ctx-doctor
mksglu/context-modeRuns diagnostics on the context-mode install: runtimes, hooks, FTS5, plugin registration and versions.
ctx-index
mksglu/context-modeIndexes a local file or folder into a searchable knowledge base so later searches avoid rereading it.
ctx-purge
mksglu/context-modePermanently deletes indexed content and session data, for one session or the whole project.
ctx-search
mksglu/context-modeSearches the indexed knowledge base for focused snippets from local project files and docs.
ctx-stats
mksglu/context-modeShows how much context window context-mode saved this session, with token use per tool.
ctx-upgrade
mksglu/context-modePulls the latest context-mode from GitHub, rebuilds it, reinstalls it, and reconfigures its hooks.
planning-with-files
OthmanAdi/planning-with-filesKeeps task plans, findings and progress in project files so multi-step agent work survives context loss.
planning-with-files-ar
OthmanAdi/planning-with-filesKeeps task_plan.md, findings.md and progress.md on disk to plan and track multi-step agent work.
planning-with-files-de
OthmanAdi/planning-with-filesKeeps task_plan.md, findings.md and progress.md on disk to plan and track multi-step agent work.
planning-with-files-es
OthmanAdi/planning-with-filesKeeps task_plan.md, findings.md and progress.md on disk to plan and track multi-step agent work.
planning-with-files-zh
OthmanAdi/planning-with-filesKeeps task_plan.md, findings.md and progress.md on disk to plan and track multi-step agent work.
planning-with-files-zht
OthmanAdi/planning-with-filesKeeps task_plan.md, findings.md and progress.md on disk to plan and track multi-step agent work.
distilly
titanwings/distillySets up an owner profile for a personal agent and distills source material into person profiles.
ce-bakeoff
EveryInc/compound-engineering-pluginBuilds several independent candidate solutions to one brief, compares them and synthesizes the strongest.
ce-handoff
EveryInc/compound-engineering-pluginWrites a session handoff for another agent, or resumes from a chosen continuity source.
ce-optimize
EveryInc/compound-engineering-pluginRuns a measured loop that attributes a workload's cost or scores variants, keeping only confirmed winners.
ce-plan
EveryInc/compound-engineering-pluginProduces a structured plan for multi-step software or non-software work, without implementing it.
ce-pov
EveryInc/compound-engineering-pluginForms a grounded, decisive verdict on an adoption question, document, or set of approaches.
sessions
slopus/happySearches past Claude Code, Codex and Cursor sessions to answer what was tried or decided before.
Confidence Check
SuperClaude-Org/SuperClaude_FrameworkScores pre-coding confidence on duplicates, architecture, docs, OSS references and root cause.
pm
SuperClaude-Org/SuperClaude_FrameworkRuns project management with PDCA cycles, confidence checks and session context restore.
token-efficiency
SuperClaude-Org/SuperClaude_FrameworkSwitches the agent to terse, abbreviated output to save tokens when context is tight.
leader
KKKKhazix/khazix-skillsTurns a rough idea into a self-contained goal brief with checks an autonomous agent can run.
neat-freak
KKKKhazix/khazix-skillsReconciles project docs, agent rule files and memory with the code so the next session starts current.
pua
tanweai/puaAfter repeated failures, pushes the agent to exhaust every fix using pressure-style motivational prompts.
pua-en
tanweai/puaAfter repeated failures, uses performance-review style pressure to make the agent exhaust all options.
pua-ja
tanweai/puaAfter repeated failures, uses Japanese-style corporate pressure to make the agent exhaust every option.
skillopt-sleep
microsoft/SkillOptRuns a sleep cycle that mines past sessions and proposes held-out-gated CLAUDE.md and SKILL.md improvements.
experiment-bridge
wanshuiyin/Auto-claude-code-research-in-sleepTurns a research experiment plan into implemented code, deploys it to GPUs and collects first results.
context-compression
muratcankoylan/Agent-Skills-for-Context-EngineeringTeaches compaction and handoff summaries that keep decisions, files, risks and next steps in long sessions.
context-degradation
muratcankoylan/Agent-Skills-for-Context-EngineeringDiagnoses context failures such as lost-in-middle, poisoning and clash, with a mitigation for each pattern.
context-engineering-collection
muratcankoylan/Agent-Skills-for-Context-EngineeringGuides context engineering, multi-agent design and evaluation for production agent systems.
context-fundamentals
muratcankoylan/Agent-Skills-for-Context-EngineeringExplains context window anatomy, attention mechanics and the U-shaped curve behind context engineering.
context-optimization
muratcankoylan/Agent-Skills-for-Context-EngineeringGuides context budgeting, observation masking, KV-cache reuse and partitioning to cut token cost.
filesystem-context
muratcankoylan/Agent-Skills-for-Context-EngineeringOffloads tool output and state to files as scratchpads and handoffs, loading context just in time.
harness-engineering
muratcankoylan/Agent-Skills-for-Context-EngineeringDesigns harnesses for autonomous agent loops with locked surfaces, durable logs, rollback and approval gates.
latent-briefing
muratcankoylan/Agent-Skills-for-Context-EngineeringGuides sharing orchestrator memory with worker agents through KV cache compaction instead of text summaries.
long-horizon-prompting
muratcankoylan/Agent-Skills-for-Context-EngineeringDesigns launch prompts for long-running autonomous or parallel agents, with exact success criteria and stop rules.
memory-systems
muratcankoylan/Agent-Skills-for-Context-EngineeringDesigns persistent agent memory covering entity tracking, temporal validity, graph or vector retrieval, and consolidation.
multi-agent-patterns
muratcankoylan/Agent-Skills-for-Context-EngineeringGuides multi-agent system design: context isolation, supervisor or swarm coordination, handoffs, and when to use them.
think
AgriciDaniel/claude-obsidianApplies a ten-stage reasoning loop to consequential decisions, reviews, and tradeoffs without editing files.
skill-builder
yusufkaraaslan/Skill_SeekersDetects the source type of a URL, repo, PDF or video and builds an AI skill from it with Skill Seekers.
ao-actions
OrchestratorInc/agent-orchestratorActs on one AO session or PR: stop, resume, end, merge, review, screenshot, send files, keep host awake.
ao-answer
OrchestratorInc/agent-orchestratorFinds AO sessions waiting on the user and sends the user's chosen answer back to that session.
ao-orchestrator
OrchestratorInc/agent-orchestratorPasses the user's request to a project's AO orchestrator, waits for its reply, and relays it back.
ao-rules
OrchestratorInc/agent-orchestratorShared rules and connection helper that every ao-* skill follows, covering AO hosts and session naming.
ao-status
OrchestratorInc/agent-orchestratorRead-only report of AO sessions across all hosts: running, waiting, idle, exited, and PRs ready to merge.
hive.colony-progress-tracker
aden-hive/hiveLets a worker agent claim tasks and record step progress in a colony SQLite queue.
hive.context-preservation
aden-hive/hiveMakes the agent copy key values from tool results into working notes before context pruning removes them.
hive.error-recovery
aden-hive/hiveMakes the agent classify tool failures and retry, fix, record or escalate each one instead of dropping it.
hive.note-taking
aden-hive/hiveMakes the agent keep a structured scratchpad of decisions, extracted values and open questions.
hive.quality-monitor
aden-hive/hiveMakes the agent periodically check its output for drift, dropped items and contradictions, then adjust.
hive.worker-delegation
aden-hive/hiveGuides the agent to split colony work across parallel workers using a tracker table and playbook.
hive.writing-hive-skills
aden-hive/hiveGuides the agent to write new Agent Skills that follow the spec, with the right scope and directory.
learn
tigerless-labs/autoharnessDistills reusable lessons from the current session and stages skill-library updates through the proposal tool.
ui-skills-root
ibelick/ui-skillsRoutes a UI task to the smallest useful set of UI skills through the ui-skills CLI before implementation.
half-clone
ykdojo/claude-code-tipsClones the later half of the current conversation into a new session to save tokens while keeping recent work.
handoff
ykdojo/claude-code-tipsWrites or updates a HANDOFF.md file with goal, progress and next steps so a fresh session can continue.
quarter-clone
ykdojo/claude-code-tipsClones the last quarter of the current conversation to cut context size while keeping recent work.
review-claudemd
ykdojo/claude-code-tipsReviews recent conversation history to suggest improvements to CLAUDE.md files.
skill-developer
diet103/claude-code-infrastructure-showcaseGuides creating Claude Code skills: trigger rules, hooks, enforcement levels and progressive disclosure.
plannotator
backnotprop/plannotatorReference for choosing Plannotator commands for plan review, code review, annotations, exports and the Inbox.
plannotator-compound
backnotprop/plannotatorAnalyzes denied plan archives for feedback patterns and writes an HTML report with prompt improvements.
plannotator-inbox
backnotprop/plannotatorOpens the Plannotator Inbox and teaches the agent to leave the person messages and questions to answer later.
plannotator-setup-goal
backnotprop/plannotatorTurns an idea into a /goal package via an interview, a reviewed fact sheet and codebase exploration.
evolve
revfactory/harnessTurns feedback from past harness runs into updates for its agents, skills and orchestrator, with a change log.
harness
revfactory/harnessDesigns a project's agent team and per-agent skills, then writes an orchestrator skill and a CLAUDE.md entry.
stitch-loop
google-labs-code/stitch-skillsRuns an autonomous baton-passing loop that builds site pages with Stitch, one iteration at a time.
workflow-runner
jnMetaCode/superpowers-zhRuns agency-orchestrator YAML multi-role workflows inside the current agent session, without an API key.
ccpm
automazeio/ccpmGuides spec-driven delivery from PRD to epic, GitHub issues and parallel agent work with status tracking.
hermes-ephemeral-delegation
Gentleman-Programming/gentle-aiTells the orchestrating agent when to delegate work to fresh, ephemeral workers and how to brief them.
rdd-advisory-transport
Gentleman-Programming/gentle-aiEnforces the shared Go contract for how coding agent runtimes invoke advisory code reviewers.
rdd-defect-workflow
Gentleman-Programming/gentle-aiGuides bounded review-driven defect work with receipts, delivery gates and approval checks before implementation.
code-improver
trailofbits/skillsRuns an autonomous review-and-fix loop on any code target until a named reviewer passes it.
skill-improver
trailofbits/skillsRuns a review-and-fix loop on a Claude Code skill until its review comes back clean.
agent-orchestration-advisor
deanpeters/Product-Manager-SkillsGuides designing multi-agent workflows with clear boundaries, handoffs and monitoring for PM tasks.
autonomous-investigation
deanpeters/Product-Manager-SkillsSets a shared protocol for autonomous research: labeled evidence, search-plan gate, and diffable outputs.
health
tw93/WazaAudits Claude, Codex, or Pi agent config, instructions, hooks, and MCP setup for drift and misconfig.
article-prompts-to-skills
MengTo/SkillsTurns an article or prompt pack into small, validated reusable agent skills with demos and a narrow commit.
audit-verify-explain-grade-5
MengTo/SkillsAudits changes, verifies claims with evidence, and explains results in plain grade-5 language.
manage-skills
xingkongliang/skills-managerInstalls, updates, removes and deploys agent skills through the skills-manager CLI.
darwin-skill
alchaincyf/darwin-skillScores SKILL.md files on a nine-point rubric, then hill-climbs edits with blind judges and git rollback.
deepchat-sdd
ThinkInAIXYZ/deepchatGuides the agent to write a durable SDD plan with an execution path before substantial DeepChat changes.
memory-management
ThinkInAIXYZ/deepchatGuides the agent to recall and store durable learning in Memory, Skills, Scheduled Tasks or Tape correctly.
audit-agents-skills
FlorianBruniaux/claude-code-ultimate-guideScores Claude Code agents, skills and commands on 16 weighted criteria and grades production readiness.
source-command-methodology-advisor
FlorianBruniaux/claude-code-ultimate-guideReads the codebase, asks three targeted questions, then recommends an AI-assisted development methodology stack.
auto
Q00/ouroborosTurns a multi-file task into a spec, then runs and verifies it in the background.
brownfield
Q00/ouroborosLists, scans and sets default existing repositories used as context for interviews.
cancel
Q00/ouroborosLists and cancels stuck or orphaned Ouroboros executions and background jobs via MCP.
interview
Q00/ouroborosRuns a Socratic interview that turns a vague request into goals, constraints and acceptance criteria.
ouroboros-help
Q00/ouroborosPrints a reference guide to Ouroboros commands, agents and the workflow loop.
ouroboros-run
Q00/ouroborosExecutes a Seed spec file as a background job and checks each acceptance criterion against the result.
ouroboros-status
Q00/ouroborosShows an Ouroboros session's status and, on request, measures its drift from the seed goal.
ralph
Q00/ouroborosRuns a background loop that repeats evolve steps until QA passes, converges, or hits its generation limit.
resume-session
Q00/ouroborosLists unfinished Ouroboros sessions after a disconnect and shows the commands to reattach to them.
seed
Q00/ouroborosTurns settled goals and constraints into a validated, immutable Seed spec with acceptance criteria.
tutorial
Q00/ouroborosWalks the user through Ouroboros hands-on, from a rough idea to interview, Seed, run and evaluate.
unstuck
Q00/ouroborosBrings five lateral-thinking personas to a stuck problem or hard choice and converges on one recommendation.
asking-user-questions
JetBrains/thinkrailShared norms for interviewing users in rounds with ask_user_question: question design, options and fallback.
choosing-a-workflow
JetBrains/thinkrailRoutes new project onboarding, PR lifecycle work and behavior changes to the matching workflow skill.
setting-up-a-project
JetBrains/thinkrailRoutes a project to the right spec-setup flow: existing specs, new project interview, or codebase import.
starting-a-new-project
JetBrains/thinkrailTurns a raw idea in an empty workspace into a goal-and-requirements doc, confirmed section by section.
todos
JetBrains/thinkrailKeeps a chat-scoped plan of grouped tasks and steps, with loose items queued at the end for later pickup.
writing-workflow-skills
JetBrains/thinkrailChecklist for adding or changing workflow skills in the pi-thinkrail-workflow package against its spec.
create-custom-agent
dotnet/skillsScaffolds VS Code .agent.md custom agent files with tool lists, instructions and handoffs.
create-skill
dotnet/skillsScaffolds new agent SKILL.md files with valid names, frontmatter, sections and directory layout.
improve-skill-quality
dotnet/skillsDiagnoses skill evaluations that regress or lack power, then suggests targeted fixes to skills or evals.
continuous-learning
WorldFlowAI/everything-claude-codeExtracts reusable patterns from finished sessions and saves them as learned skills via a Stop hook.
strategic-compact
WorldFlowAI/everything-claude-codeSuggests manual context compaction at task boundaries, triggered by a tool-call count hook.
next-idea
breaking-brake/cc-wf-studioRuns one unattended ideation pass that proposes user-facing improvements as idea issues, without coding.
next-qa
breaking-brake/cc-wf-studioRuns one unattended QA iteration: updates open QA PRs, then builds one queued test task as a PR.
next-qa-idea
breaking-brake/cc-wf-studioFinds the most valuable untested behavior and files one qa issue describing the test, without writing code.
next-task
breaking-brake/cc-wf-studioRuns one unattended implementation pass: fixes urgent interrupts, or builds one queued idea into a PR.
customize
aipoch/open-scienceRoutes requests to create or manage Specialist agents and Skills on the Open Science platform.
avoid-ai-writing-router
conorbronsdon/avoid-ai-writingRoutes a multi-step AI-writing task (audit, rewrite, file edit, verify) to the right sub-skill in order.
omh-adversarial-consensus
rlaope/oh-my-hermesHas independent perspectives attack a technical proposal, then distills the findings for a separate planner.
omh-agent-debug
rlaope/oh-my-hermesDiagnoses a stuck, looping, or drifting agent run and prepares the smallest safe recovery step.
omh-agent-instructions
rlaope/oh-my-hermesWrites or updates AGENTS.md or CLAUDE.md with only what the code cannot show, and verifies commands.
omh-agent-ops-review
rlaope/oh-my-hermesSummarizes AI agent progress, blockers, quality gates and throughput for a manager's review.
open-knowledge-write-skill
inkeep/open-knowledgeGuides the agent through writing a new SKILL.md, from scoping and frontmatter to structure and installation.
council
0xNyk/council-of-high-intelligenceConvenes historical-thinker personas to debate a problem from several angles and reach a verdict.
adhd
UditAkhourii/adhdGenerates design options in parallel from several thinking frames, then scores, prunes and deepens the best.
flow-deliver
nyldn/claude-octopusValidates, scores and reviews work using whichever external AI providers are available.
flow-develop
nyldn/claude-octopusImplements a task by coordinating available external AI providers in the Develop phase.
flow-discover
nyldn/claude-octopusResearches a topic by dispatching background probes to external AI providers and synthesizing results.
flow-parallel
nyldn/claude-octopusSplits a large change or migration into work packages run in parallel, each with quality gates.
octopus-quick
nyldn/claude-octopusHandles small ad-hoc tasks directly, skipping the full multi-phase workflow.
agent-context-isolation
parcadei/Continuous-Claude-v3Sets rules for background agents that write results to files so sub-agent transcripts stay out of the main context.
agent-orchestration
parcadei/Continuous-Claude-v3Tells the agent to hand multi-file implementation work to sub-agents so the main context stays small.
autonomous-builder
foryourhealth111-pixel/Vibe-SkillsRuns a full software lifecycle autonomously: spec, implementation, tests, debugging, and deployment.
vibe
foryourhealth111-pixel/Vibe-SkillsRuns tasks in a governed mode that freezes requirements, bounds execution, and enforces verification.
cc-sdd-new-agent
gotalab/cc-sddMakes the agent add a new coding-agent integration to cc-sdd by following its plan-first SOP.
kiro-discovery
gotalab/cc-sddEntry point that picks the right spec path for new work and refines ideas through structured questions.
kiro-impl
gotalab/cc-sddImplements approved spec tasks with TDD, optionally dispatching a fresh sub-agent per task with review.
kiro-review
gotalab/cc-sddAdversarially reviews a finished task against its spec, boundaries and verification evidence before approval.
kiro-spec-batch
gotalab/cc-sddGenerates requirements, design and tasks for every pending roadmap feature, running independent ones in parallel waves.
kiro-spec-design
gotalab/cc-sddWrites a technical design for a spec, mapping each requirement to components, interfaces and diagrams.
kiro-spec-init
gotalab/cc-sddInitializes a new spec folder with spec.json and a requirements template, clarifying the feature intent first.
kiro-spec-quick
gotalab/cc-sddRuns init, requirements, design and tasks in one pass, interactively or automatically with --auto.
kiro-spec-requirements
gotalab/cc-sddGenerates EARS-format requirements for a spec, with explicit scope boundaries and a pre-write review gate.
appllama-usage
Appllama/appllama-skillsTeaches the agent to research real top app screens via the Appllama MCP, then build from them.
task-observer
rebelytics/one-skill-to-rule-them-allLogs task friction and user corrections to a workspace log, then turns them into skill updates.
architect
michael-denyer/pstack-claudeSketches types, signatures and module layout before coding, then implements against the sketch.
arena
michael-denyer/pstack-claudeRuns parallel candidate attempts on one task, picks a base, and grafts the strongest parts together.
automate-me
michael-denyer/pstack-claudeTurns the user's working conventions and recent transcripts into a personal -mode skill for agents.
correct
michael-denyer/pstack-claudeTurns repeated agent mistakes in a repo into architecture, type, lint, or test checks that catch them.
next
agenticnotetaking/arscontextaRecommends one next action from task, queue, inbox and health state, without executing it.
remember
agenticnotetaking/arscontextaSaves workflow friction from your corrections or sessions as methodology notes for later reuse.
rethink
agenticnotetaking/arscontextaReviews logged observations and tensions to test system assumptions and propose changes.
claude-md-review
wesammustafa/Claude-Code-Everything-You-Need-to-KnowAudits a CLAUDE.md for vague rules, unnamed files, stale facts and bloat, with rewrite suggestions.
onboard
wesammustafa/Claude-Code-Everything-You-Need-to-KnowExplains a repo's shared Claude Code setup to a new teammate without changing any files.
stop-that-shit
lennney/stop-that-shitKeeps the agent from speculative hardening and scope creep while finishing the requested work fully.
claudex-loop
chaseai-yt/claudex-loopHardens a plan through Claude and Codex review rounds, then optionally builds and cross-inspects it.
claudex-route
chaseai-yt/claudex-loopRecommends which model or agent should handle a task, and can run one scoped handoff on request.
codex-build
chaseai-yt/claudex-loopHave Codex build a spec-defined work order, then have Claude independently inspect the changes.
grill-me-codex
chaseai-yt/claudex-loopInterviews you about a plan, then has Codex review it read-only until approved, before any code.
grill-with-docs-codex
chaseai-yt/claudex-loopInterviews you against project docs and glossary, updates them, then has Codex review the plan.
loop-library
Forward-Future/loopyLegacy alias for Loopy: discovers, audits, runs and debriefs repeatable, bounded AI agent loops.
loopy
Forward-Future/loopyDiscovers, audits, crafts, runs and debriefs repeatable AI agent loops, each with bounded terminal states.
vibe-agents
KhazP/vibe-coding-prompt-templateWrites stable project agent instructions and tool config from agreed product and technical decisions.
vibe-handoff
KhazP/vibe-coding-prompt-templateWrites a compact, factual project handoff for another session or tool without transcripts or secrets.
code-task-generator
mikeyobrien/ralph-orchestratorGenerates structured .code-task.md files with Given-When-Then criteria from descriptions or PDD plans.
find-code-tasks
mikeyobrien/ralph-orchestratorLists all code task files in the repo with their status, dates and metadata, with optional status filters.
ralph-docs
mikeyobrien/ralph-orchestratorAnswers Ralph Orchestrator questions from its published docs before guessing, and scopes code changes.
ralph-hats
mikeyobrien/ralph-orchestratorCreates, validates and explains Ralph hat collection files in .ralph/hats for multi-hat workflows.
ralph-loop
mikeyobrien/ralph-orchestratorStarts, monitors, resumes, merges and debugs Ralph loops using the ralph CLI and loop state files.
breezing
Chachamaru127/claude-code-harnessRuns Plans.md tasks as a Codex team, as an alias of harness-work with optional Cursor worker delegation.
cursor-ask
Chachamaru127/claude-code-harnessSends read-only questions and design reviews to Cursor's agent in ask mode, which cannot write files.
cursor-do
Chachamaru127/claude-code-harnessHands one file edit to Cursor Composer in a separate worktree, then reviews the diff before merging.
harness-loop
Chachamaru127/claude-code-harnessRuns a background Codex loop that executes ready task batches via Breezing, with status and stop controls.
harness-work
Chachamaru127/claude-code-harnessExecutes Plans.md tasks end to end with Codex-native tools, from one task to a full parallel team run.
figure-it-out
backnotprop/pstackDesigns an auditable, phased playbook for large or unfamiliar tasks and logs each decision.
continual-learning
microsoft/skillsGuides setting up hooks and two-tier memory so coding agents learn from past sessions.
create-agent
open-gitagent/opengapCreates and configures agent.yaml, SOUL.md and skill folders for a gitagent agent directory.
agent-teams
rohitg00/pro-workflowCoordinates several Claude Code sessions as a team with shared task lists, messaging and file locks.
auto-setup
rohitg00/pro-workflowDetects a project's type and writes Claude Code settings with lint, typecheck, test and format gates.
batch-orchestration
rohitg00/pro-workflowSplits large repo-wide changes into independent units and runs parallel agents in isolated worktrees.
compact-guard
rohitg00/pro-workflowSaves task state and key files before context compaction, then restores them afterward.
context-optimizer
rohitg00/pro-workflowGuides context and token budget management: when to compact, trim tools, and delegate heavy output.
cost-tracker
rohitg00/pro-workflowHelps check session costs, set budget limits, and cap tool calls to control token spend.
inbox-triage
nano-muse/nanoMuseSorts recent mail into reply, waiting, to-do, read-later and ignore groups, leaving drafts unsent.
meeting-prep
nano-muse/nanoMuseBuilds a one-page brief for a meeting: attendees, last exchanges, open items and related files.
llm-council
aiwithremy/claude-skills-llm-councilSends a decision to five advisors who review each other anonymously, then a chairman gives a verdict.
memory-config
zilliztech/memsearchDiagnoses and changes MemSearch memory settings such as providers, summaries, directories and index health.
memory-recall
zilliztech/memsearchSearches past session memories via memsearch and returns the few chunks relevant to the current question.
memory-to-skill
zilliztech/memsearchDistills repeated workflows from MemSearch memory into candidate skills, reviewed before any install.
find-skills
heshengtao/super-agent-partyFinds and installs agent skills from the open skills ecosystem when a user asks how to do a task.
yao-meta-skill
yaojingang/yao-meta-skillCreates, improves and evaluates agent skills from workflows, prompts and SOPs, with release checks.
agent-md-refactor
softaworks/agent-toolkitGuides the agent to split a bloated AGENTS.md or CLAUDE.md into a short root file and linked docs.
command-creator
softaworks/agent-toolkitGuides the agent to create reusable Claude Code slash command files with clear structure and steps.
agy-delegate
amElnagdy/delegate-skillsHands a bounded coding task to the Antigravity agy CLI, then reviews its diff before landing it.
aider-delegate
amElnagdy/delegate-skillsHands a bounded coding task to Aider, then reviews its diff before landing it yourself.
claude-delegate
amElnagdy/delegate-skillsHands a bounded coding task to a separate Claude Code session, then reviews its diff before landing it.
cline-delegate
amElnagdy/delegate-skillsHands a bounded coding task to the Cline CLI, then reviews its diff before landing it yourself.
codex-background
amElnagdy/delegate-skillsLets a Codex orchestrator wait on delegate runs in the background through an MCP relay, without polling.
codex-delegate
amElnagdy/delegate-skillsHands a bounded coding task to the Codex CLI, then reviews its diff before landing it yourself.
commandcode-delegate
amElnagdy/delegate-skillsHands a bounded coding task to the Command Code CLI, then reviews its diff before landing it yourself.
copilot-delegate
amElnagdy/delegate-skillsHands a bounded coding task to the GitHub Copilot CLI as a background implementer, then reviews its diff.
cursor-delegate
amElnagdy/delegate-skillsHands a bounded coding task to the Cursor Agent CLI as a background implementer, then reviews its diff.
delegate-setup
amElnagdy/delegate-skillsMaps installed implementer CLIs to task lanes and writes the config only after the user approves.
learning-opportunities
DrCatHicks/learning-opportunitiesOffers brief optional learning exercises after architectural work to deepen understanding of design choices.
full-send
t3dotgg/tstackDrives a task autonomously to a merged PR, with permission to merge once the merge gate passes.
full-stop
t3dotgg/tstackCompletes a task autonomously to a ready, unmerged PR, leaving the final merge to the user.
agent-skills-platform
FrancyJGLisboa/agent-skills-platformCreates, validates, exports and audits cross-platform agent skills from a described workflow.
agents-best-practices
DenisSergeevitch/agents-best-practicesGuides designing, auditing, and debugging agent harnesses across providers, covering loops, tools, memory, and evals.
fable-domain
Sahir619/fable-methodBuilds a researched domain skill bundle: flowchart workflow, domain adapter, trap fixture, smoke eval.
fable-loop
Sahir619/fable-methodRuns a task through planning, parallel evidence gathering, execution, adversarial verification, and an outcome report.
fable-method
Sahir619/fable-methodMakes the agent classify a task, define done, gather evidence, act surgically, verify, and report outcome first.
evolver
codejunkie99/agentic-stackDiagnoses agent failures, then tests prompt or config fixes against a paired screen before installing them.
huashu-agent-swarm
alchaincyf/huashu-skillsRuns parallel Claude agents in tmux on git worktrees, coordinating through git and task lock files.
dynamic-workflow
pchalasani/claude-code-toolsWrites and runs durable JavaScript workflows that coordinate multiple headless Codex agents in staged parallel work.
github-wake
pchalasani/claude-code-toolsRegisters an existing GitHub issue so the current session resumes when the next comment arrives.
msg
pchalasani/claude-code-toolsHas the agent register, send, reply to and read messages with other coding agent sessions via the msg CLI.
recover-context
pchalasani/claude-code-toolsRecovers the last task's context from the parent session using a sub-agent to protect the main context.
session-search
pchalasani/claude-code-toolsSearches earlier Claude Code or Codex sessions for past work, decisions, or code patterns and summarizes them.
autonomous-loops
xu-xiang/everything-claude-code-zhPatterns for unattended Claude Code loops, from simple claude -p pipelines to multi-agent DAG orchestration.
configure-ecc
xu-xiang/everything-claude-code-zhWalks you through choosing and installing ECC skills and rules into user or project Claude Code folders.
continuous-learning-v2
xu-xiang/everything-claude-code-zhObserves sessions through hooks, saves confidence-scored instincts, and evolves them into skills, commands or agents.
create-agent-skills
glittercowboy/taches-cc-resourcesGuides the agent in creating, structuring and refining Claude Code skills as SKILL.md files.
create-hooks
glittercowboy/taches-cc-resourcesGuides the agent in writing Claude Code hooks that run commands or prompts on tool and session events.
create-plans
glittercowboy/taches-cc-resourcesMakes the agent write hierarchical, executable project plans with verification criteria for solo work.
create-slash-commands
glittercowboy/taches-cc-resourcesMakes the agent create Claude Code slash commands with YAML frontmatter, XML structure and arguments.
create-subagents
glittercowboy/taches-cc-resourcesMakes the agent create Claude Code subagents with focused prompts, tool limits and Task tool delegation.
setup-ralph
glittercowboy/taches-cc-resourcesMakes the agent set up an autonomous Ralph coding loop with prompts, structure and test backpressure.
init
mcollina/skillsCreates or prunes an AGENTS.md file with minimal repo instructions agents cannot discover themselves.
skill-optimizer
mcollina/skillsOptimizes AI skill packs for activation, clarity and cross-model reliability using benchmark loops.
claude-reflect
BayramAnnakov/claude-reflectCaptures user corrections during sessions and reminds you to run /reflect to update CLAUDE.md.
actualize
NeoLabHQ/context-engineering-kitSyncs the .fpf knowledge base with recent git changes, flagging stale evidence and context drift.
add-task
NeoLabHQ/context-engineering-kitCreates a draft task file in .specs/tasks/draft/ that preserves the user's original intent.
agent-evaluation
NeoLabHQ/context-engineering-kitEvaluates Claude Code commands, skills and agents with outcome-focused rubrics to guide improvements.
apply-anthropic-skill-best-practices
NeoLabHQ/context-engineering-kitApplies Anthropic's official skill authoring guidance to structure and tighten a complex SKILL.md.
grill-me
nateherkai/AIS-OSInterviews the user about a plan or design, saving every answer to a brainstorm file as it goes.
level-up
nateherkai/AIS-OSRuns a weekly interview that picks one workflow to automate and ships a single verified artifact or repair.
link
nateherkai/AIS-OSAdds a concise routing entry for a file, folder, or URL into AGENTS.md or the AIOS index.
mex-relay
mex-memory/mexPrepares a durable, git-native handoff note so another engineer can pick up the work where it stopped.
unreal-skill
EpicGames/unreal-engine-skills-for-claude-code-pluginGuides writing and reviewing Unreal Engine Agent Skills that the in-editor agent loads through unreal-mcp.
thinking-bounded-rationality
tjboudreaux/cc-thinking-skillsSets a good-enough threshold before searching, then stops at the first option that meets it.
thinking-circle-of-competence
tjboudreaux/cc-thinking-skillsChecks whether a claim has grounding before asserting it, then answers, fetches, or abstains.
thinking-cynefin
tjboudreaux/cc-thinking-skillsClassifies a problem's cause-effect domain first, then picks the matching response approach.
thinking-effectuation
tjboudreaux/cc-thinking-skillsUnder uncertainty, the agent starts from its means and affordable loss, then acts for real commitments.
thinking-first-principles
tjboudreaux/cc-thinking-skillsWhen a constraint looks fixed, the agent reduces it to verified primitives and rebuilds a simple fix.
checkpoint
pedrohcgs/claude-code-my-workflowSaves the active plan, decisions, file pointers and next steps to a snapshot file for resuming later.
compress-session
pedrohcgs/claude-code-my-workflowDistills the current session into a structured note of decisions, open questions and next steps.
codex-skill
feiskyer/claude-code-settingsDelegates coding, code review and plan review to the OpenAI Codex CLI in sandboxed modes.
agent-communication
codeaholicguy/ai-devkitLets the agent list, inspect and message other active Codex or Claude Code agents via ai-devkit.
agent-management
codeaholicguy/ai-devkitLets the agent start, inspect, message, group, resume and stop running agents via ai-devkit.
dev-implementation
codeaholicguy/ai-devkitGuides implementation against plan and design using TDD and verification before tasks are marked complete.
hivemind-memory
activeloopai/hivemindSearches the team's shared Activeloop memory index and summaries alongside built-in memory.
agent-team-design
agentlas-ai/Agentlas-OSDesigns a multi-agent team's roles, handoffs, memory and policy owners, and visible agents folder layout.
agentlas-auto-activation
agentlas-ai/Agentlas-OSAdds or audits project memory files and opt-in activation rules that make a folder Agentlas-aware.
agentlas-build
agentlas-ai/Agentlas-OSGuides designing, assembling and packaging a single agent or multi-agent team through its build workflow.
minutes-live-sidekick
silverstein/minutesAct as a live meeting sidekick in the terminal: answer questions, flag risks, track decisions.
agmsg
fujibee/agmsgSend messages between local coding agents such as Claude Code and Codex through a SQLite store.
agent-team
dyoshikawa/rulesyncRuns an implementer and reviewer subagent loop until the review has no high findings and at most three mid.
batch-all-issues
dyoshikawa/rulesyncWorks through every open issue one at a time, closing those that need no action and fixing and merging the rest.
batch-all-scrap-issues
dyoshikawa/rulesyncWorks through every open maintainer-scrap issue one at a time, fixing, merging or closing each.
handoff-protocols
rohitg00/skillkitWrites structured handoff documents so another person or agent can take over a task.
Create New Skills
disler/claude-code-hooks-multi-agent-observabilityGuides the agent to write a new Claude Code skill with correct frontmatter and progressive disclosure.
attention-kind
alexgreensh/attention-spanSwitches replies to an ADHD-friendly style: short, scannable, with the key point first, for the rest of the chat.
rundown
alexgreensh/attention-spanSwitches replies to a briefing format: a TL;DR, a status checklist, and numbered choices for next steps.
spartan
alexgreensh/attention-spanSwitches replies to a terse, no-filler style that leads with the answer, for the rest of the chat.
gentle-ai
Gentleman-Programming/gentle-shellApplies a senior-architect harness to Pi coding work: clarify scope first, test-first changes, and delegation rules.
discover
evo-hq/evoSets up evo in a repo by proposing optimization dimensions, building a benchmark and running a first experiment.
optimize
evo-hq/evoRuns the evo optimization loop, with the orchestrator planning rounds and subagents running experiments.
report
evo-hq/evoSummarizes an evo run from stored state only, without running benchmarks or experiments.
subagent
evo-hq/evoProtocol evo subagents follow: a structured brief, verifier checks, and required result reporting.
building-with-ai-agents
RefoundAI/lenny-skillsHelps delegate well-defined coding tasks to parallel AI agents with specific prompts and reviews.
communicating-concisely
GanyuanRan/AegisSwitches the agent to terse caveman-style replies with fewer tokens until the user turns it off.
first-principles-review
GanyuanRan/AegisRuns a light first-principles check before direction-setting choices to curb fallback growth.
goal-framing
GanyuanRan/AegisMakes the agent write a short goal frame with success evidence and stop condition before starting work.
long-task-continuation
GanyuanRan/AegisKeeps multi-step or multi-session tasks checkpointed and resumable, with drift checks and evidence gates.
nopua
wuji-labs/nopuaPushes the agent to keep working through repeated failures with a structured recovery method instead of giving up.
nopua-lite
wuji-labs/nopuaA compact version of the recovery method: after failures, pause, list attempts, then try new angles.
nopua-zh
wuji-labs/nopuaAfter repeated failures, makes the agent try more methods, verify results and report evidence.
skill-upper
alibaba/skill-upRuns, diagnoses and iterates on Agent Skill evaluations using the skill-up command line tool.
audit-agents
dzhng/skillsAudits or rewrites AGENTS.md so it keeps only lasting principles and moves mechanics to their owning docs.
audit-decisions
dzhng/skillsTraces the unstated decisions an AI implementer made into a verdict ledger, without changing any code.
auto-research
dzhng/skillsRuns staged experiments from a hypothesis queue to map parameter effects and keep verified best artifacts.
claude
dzhng/skillsRuns Claude Code as an independent second-agent subagent for opinions or delegated tasks, with model and effort options.
close-spec
dzhng/skillsArchives a finished spec and rewrites it from a build plan into a rationale record.
eval-skills
dzhng/skillsTests a skill blind against golden input-output cases in fresh subagents, then edits the skill to close the gaps.
write-skills
dzhng/skillsGuides creating or revising agent skills, covering trigger descriptions, brevity, and progressive disclosure.
better-codex
noobnooc/agentAdds behavioral guardrails that keep coding scope tight, verify fixes, and avoid shallow patches.
lore
specstoryai/getspecstoryMines past AI coding session histories for repeated workflows and turns them into installable skills.
workthreads
specstoryai/getspecstoryBuilds a weekly rollup of work threads from SpecStory coding histories, marked new, open, or recently closed.
oma-brainstorm
first-fluke/oh-my-agentExplores goals, constraints, and alternative approaches, then saves an approved design before planning.
oma-coordination
first-fluke/oh-my-agentGuides step-by-step coordination of multi-domain specialist tasks, with handoffs and contract checks.
deja-history
vshulcz/deja-vuSearches the user's past AI coding sessions to reuse earlier decisions, rules and fixes before editing.
deja-search
vshulcz/deja-vuSearches the user's past AI coding sessions with the deja CLI to reuse earlier decisions, rules and fixes.
daily-journal
huytieu/COG-second-brainHas the agent append short work log entries to a daily journal file and run guided reflections.
save-learning
caliber-ai-org/ai-setupSaves a user's stated rule or preference as a persistent project learning via the caliber CLI.
scoring-checks
caliber-ai-org/ai-setupAdds a deterministic config-quality scoring check in src/scoring/checks using shared point constants.
setup-caliber
caliber-ai-org/ai-setupInstalls Caliber and a pre-commit hook that keeps CLAUDE.md, Cursor and Copilot configs in sync on each commit.
agent-context-audit
AI-Builder-Club/skillsAudits a repo's agent context files, docs and skills against Claude 5 guidance, then applies approved fixes.
new-loop
AI-Builder-Club/skillsCreates a recurring work loop in a file-based knowledge base with a charter, then verifies it with one real run.
open-agent-teams
AI-Builder-Club/skillsRuns any CLI agent in a detached tmux session and waits for its done-signal and result file.
setup-codebase-harness
AI-Builder-Club/skillsSets up repo docs, a one-command dev stack, e2e tests and verify loops so agents can build and check changes.
skill-authoring
uvwt/agentdockGuides writing, revising and testing AgentDock skills around a portable SKILL.md core.
skill-installation
uvwt/agentdockReviews a skill's security and frontmatter, then installs, updates or removes it in AgentDock.
skill-creator
AIPexStudio/AIPexGuides the agent through creating or updating AIPex skills with SKILL.md structure and bundled resources.
openspec-apply-change
MemeMeow-Studio/MemeMeowGuides the agent through implementing the tasks of an OpenSpec change, using the openspec CLI.
openspec-archive-change
MemeMeow-Studio/MemeMeowArchives a finished OpenSpec change and merges its delta specs into the main specs.
openspec-explore
MemeMeow-Studio/MemeMeowPuts the agent in a think-only mode to explore ideas and requirements without writing code.
openspec-propose
MemeMeow-Studio/MemeMeowCreates an OpenSpec change with proposal, specs, design and tasks in one step, without editing code.
openspec-sync-specs
MemeMeow-Studio/MemeMeowMerges a change's delta specs into the main specs by direct edits, without archiving the change.
openspec-update-change
MemeMeow-Studio/MemeMeowRevises an OpenSpec change's planning documents so they stay consistent, without touching code.
grill-me
RobMitt/grill-me-skillMakes the agent interview the user one question at a time about a plan until the design is settled.
os-ask-simple
kharmanskyi/open-stepsRewrites technical questions in plain words, screens options with six checks, and gives one recommendation.
os-done-or-not
kharmanskyi/open-stepsReports whether the agent's work is finished, with proof for each claim and a clear verdict.
os-step-by-step
kharmanskyi/open-stepsTurns a request for the user to act into single-line, plain-language steps, then checks the result.
os-what-could-go-wrong
kharmanskyi/open-stepsRuns a premortem on a pending decision in a fresh agent and gives a go or no-go verdict.
os-whats-next
kharmanskyi/open-stepsPicks the next task from the last report, local changes and open PRs, and merges ready PRs.
jev-compaction
kerpopule/hermes-jev-skillsHas the agent mark each transcript turn as keep, summarize or drop when cutting it to a fixed size.
jev-frontier-work
kerpopule/hermes-jev-skillsGuides choosing a paid frontier model for hard tasks, then monitoring the delegated run for decisions.
skillpack-creator
CreminiAI/skillpackTurns a successful task into a reusable SkillPack with skills, prompts, start scripts and an optional zip.
adhd-output-style
fcakyon/claude-codex-settingsFormats replies as short numbered steps ending in one small next action, for low-friction reading.
ctx
ctxrs/ctxHas the agent search prior agent sessions and code history with ctx to recover decisions and provenance.
vc-agent-strategy-compare
withkynam/vibecode-pro-max-kitScores four execution strategies for a phase or fan-out task and recommends one with agent count math.
vc-context-discovery
withkynam/vibecode-pro-max-kitFinds and loads the project context files relevant to a task, routed by domain at session start.
spec
JuliusBrussee/cavekitCreates and amends a SPEC.md project spec, recording goals, invariants and bugs in a fixed format.
ask-questions-if-underspecified
MoizIbnYousaf/ai-agent-skillsAsks a small set of clarifying questions before implementing when a request is underspecified.
curate-a-team-library
MoizIbnYousaf/ai-agent-skillsGuides building a managed team skills library by mapping work to shelves and browsing before curating.
manage-local-skills
TencentCloudBase/CloudBase-AI-ToolkitConverts local prompt folders into standard SKILL.md skills and syncs them into agent skill directories.
planning-workflows
TencentCloudBase/CloudBase-AI-ToolkitChooses between spec-based and direct planning modes, then follows the matching requirements, design, and task steps.
repo-onboarding
hoangsonww/Claude-Code-Agent-MonitorGuides the agent through reading this repo's architecture, locating the affected layer and choosing verification commands first.
money-learn
iamzifei/show-me-the-moneySaves, searches and prunes small project learnings that money skills auto-load in later sessions.
money-ops
iamzifei/show-me-the-moneySchedules and runs business tasks like posting, outreach, reporting and health checks without manual steps.
checkpoint-mode
asklokesh/loki-modeMakes the agent pause for human review every N tasks, working in short autonomous bursts.
environment-discovery
vstorm-co/pydantic-deepagentsMakes the agent systematically inspect files, tools and existing code before acting in an unfamiliar environment.
loki-mode
asklokesh/loki-modeRuns an autonomous build-and-verify loop from a spec and ends with a signed evidence receipt.
sno-mem-claude
sno-ai/sno-stationLets Claude search, read, store and correct Sno memory through four explicit shell commands.
sno-mem-codex
sno-ai/sno-stationLets Codex search, read, store and correct Sno memory through four explicit shell commands.
sno-mem-cursor
sno-ai/sno-stationLets Cursor search, read, store and correct Sno memory through four explicit terminal commands.
decision-prompt-builder
zhnnky329/MathModeling-skillsBuilds a compact choice card so the human decides key modeling trade-offs at real judgment points.
agent-relay
AgentWorkforce/relayLets Codex coordinate sub-agents through Agent Relay messaging, handoffs, and shared status signals.
choosing-swarm-patterns
AgentWorkforce/relayHelps pick among 24 multi-agent swarm patterns, such as fan-out, pipeline, and debate, and how to run them.
creating-agent-skills
AgentWorkforce/relayGuides writing SKILL.md agent skill packages for Codex, Copilot, and Amp, with frontmatter and layout rules.
relay-80-100-workflow
AgentWorkforce/relayGuides agent workflows that run tests, fix failures, and verify features end-to-end before committing.
review-fix-signoff-loop
AgentWorkforce/relayGuides writing Agent Relay workflows that loop review and fixes until Claude and Codex both sign off.
writing-relayflows
AgentWorkforce/relayWrites Relayflows v2 flows in TypeScript or YAML, with steps, verification gates, and check/run/resume.
gastown
numman-ali/n-skillsRuns the Gas Town multi-agent system through gt and bd commands to track work with convoys and agents.
orchestration
numman-ali/n-skillsSplits complex work into parallel subtasks, tracked with persistent task dependencies and session todos.
migliora
dagoSte/squadraImproves a project's workflow based on observed failures, with checked fixes and local commits.
squadra
dagoSte/squadraCoordinates tasks, proofs and reviews across several Codex, Claude or Cursor windows via a local registry.
deep-discuss
zhu1090093659/spec_driven_developGuides a multi-phase discussion: review the problem first, then analyze, compare options and recommend.
find-skills-combo
agentscope-ai/OpenJudgeSplits a complex task into subtasks and recommends skill combinations for quality or fewest installs.
spec-driven-develop
zhu1090093659/spec_driven_developPlans large rewrites or migrations through analysis, task breakdown and tracking docs, then executes them.
ss-learn
bitjaru/styleseedSaves an approved UI design correction as a local candidate rule, with an opt-in privacy-minimized share.
self-learning
Kulaxyz/self-learning-skillsCaptures a proven workflow from the current session as a reusable Agent Skill for future sessions.
create-plan
freestylefly/wesightTurns a coding task request into one concise, ordered action plan without editing any files.
archon
SethGammon/CitadelDrives multi-session campaigns by splitting work into phases, delegating to sub-agents, and saving state.
autopilot
SethGammon/CitadelProcesses pending intake items through research, plan, build and verify, then packages them as a delivery.
skillforge
tripleyak/SkillForgeRoutes requests to find, improve, create, or audit agent skills, and maintains the skill ecosystem.
agent-factory
alirezarezvani/claude-code-skill-factoryGenerates Claude Code agent and sub-agent files with frontmatter, tool access and execution patterns.
claude-md-enhancer
alirezarezvani/claude-code-skill-factoryAnalyzes, generates and improves CLAUDE.md files to match a project's stack, structure and team size.
double-check
mrtooher/fable-modeRuns a fresh checker verification pass on finished deliverables and fixes any failures before delivery.
execution-guardrails
mrtooher/fable-modeApplies always on rules: verify problems before flagging them, batch minor warnings, and make careful find and replace edits.
fable-fable
mrtooher/fable-modeRoutes a task to a staged orchestrator agent running on Claude Fable 5.1 that delegates work and checks results.
fable-haiku
mrtooher/fable-modeRuns a staged, verified workflow on a Haiku worker agent for low cost bulk mechanical tasks.
fable-mode
mrtooher/fable-modeMakes the agent plan large tasks in stages, verify each stage with checks, and self review before delivering.
decide
coreyhaines31/makerskillsGuides a decision through targeted questions, records the call and rationale, and sets a revisit date.
blind-spot-pass
sangrokjung/claude-forgeSurfaces what you don't know before work in an unfamiliar domain, teaching key terms and questions first.
cc-dev-agent
sangrokjung/claude-forgeGuides spec-driven Claude Code work with context management, subagent dispatch and a post-dev workflow.
charter
sno-ai/sno-station-skillsWrites or reviews a charter: deliverables, out-of-bounds areas, owner decisions and provable success checks.
agent-notifications
777genius/agent-notificationsSends a desktop notification for requests, attention needs or milestones, avoiding duplicate completion alerts.
codeman
Ark0N/CodemanLets the agent start, prompt, wait on and read other CLI sessions through the Codeman HTTP API.
hire
markfulton/ai-employeesCopies one of eight open source AI employee folders into a local directory and runs its self tests.
improve
gadicc/yahoo-finance2Audits a codebase read-only and writes prioritized plans another agent can execute without context.
agy-jobs
keli-wen/agy-staffCollects, checks, cancels and follows up background agy jobs, and sets up the agy CLI.
agy-lead
keli-wen/agy-staffCoordinates an ongoing task by delegating work to AGY while the lead keeps decisions, review and delivery.
agent-self-eval
coco-research/cocoScores an agent's own output on correctness, clarity, actionability and conciseness after a run.
claude-smart
ReflexioAI/claude-smartRuns claude-smart plugin commands such as show, learn and dashboard in Codex, which has no slash commands.
fractal
plasma-ai/fractalLaunches autonomous agent loops in git worktrees, where nodes spawn child nodes to run subtasks in parallel.
memory
plasma-ai/fractalSets the discipline for a fractal node's persistent memory: topical pages, no history, clean commits.
radio
plasma-ai/fractalGuides the agent to read, send and track inter-node radio messages in a disciplined way.
storybloq
Storybloq/storybloqOpens the local project's Storybloq board and summarizes progress, blockers and handovers.
astra-flash-orchestrator
ethanplusai/codex-on-crackPlans large multi-file builds with Astra and delegates coding to a Flash worker via Codex.
crack
ethanplusai/codex-on-crackLeads a build with one configured worker and final acceptance, honoring the model you name for delegation.
crack-panel
ethanplusai/codex-on-crackShows codex-on-crack run and review status and submits screenshot evidence for visual review.
crack-plan
ethanplusai/codex-on-crackRecommends a lead and worker setup from your configured models, then lets you decide before work starts.
crack-setup
ethanplusai/codex-on-crackSets up codex-on-crack worker models and role files while preserving your lead and existing config.
generate-ai-rules
divar-ir/ai-doc-genGenerates CLAUDE.md, AGENTS.md and Cursor rule files from a repo's codebase analysis.
pause-and-resume
dongshuyan/compass-skillsPauses unfinished work at a safe point, writes a continuation checkpoint, and resumes from it when asked.
run-history-skill-builder
dongshuyan/compass-skillsTurns a completed task or workflow history into a new reusable skill package or a reviewed design plan.
chisle
JayPokale/ChislePuts the agent in terse, zero-fluff mode and applies YAGNI and a reuse-first ladder to code changes.
flow-next-audit
gmickel/flow-nextChecks .flow/memory entries against the codebase and decides keep, update, merge, replace, delete or harden.
flow-next-capture
gmickel/flow-nextTurns the current conversation into a spec file written through flowctl, with source tags on inferred content.
flow-next-chart
gmickel/flow-nextGuides one unclear idea through decision-by-decision discovery into a briefing for spec capture.
flow-next-deps
gmickel/flow-nextShows the spec dependency graph, blocking chains and parallel execution phases using flowctl.
coding-agent
swarmclawai/swarmclawDelegates coding tasks to external agents like Claude Code, Codex, Pi, or OpenCode through the shell.
best-skills
LinklyAI/best-skillsLooks up daily cross-platform rankings of agent skills to recommend or compare them, without installing.
birdview-compatibility
Qiuner/birdviewAudits installed agent skills for duplicates, broad triggers and instruction conflicts, then writes a report.
nw-ab-critique-dimensions
nWave-ai/nWaveProvides review dimensions for checking agent definitions on template compliance, size, safety and testing.
nw-abr-critique-dimensions
nWave-ai/nWaveProvides the same agent review dimensions for checking agent definitions, used by the builder-reviewer agent.
nw-agent-creation-workflow
nWave-ai/nWaveGuides the agent through a 5-phase process to design, build, validate and refine new agents.
abide-compile
coldteadotai/abideTurns a repo's AGENTS.md and CLAUDE.md rules into a checkable rubric, then validates and calibrates it.
hcom-agent-messaging
aannoo/hcomSets up hcom so AI coding agents in separate terminals can message, watch and spawn each other.
customise-workflow
anombyte93/prd-taskmasterAsks plain-language questions and writes the prd-taskmaster workflow preferences to a config file.
execute-fleet
anombyte93/prd-taskmasterRuns TaskMaster tasks in parallel git worktrees, verifies each, and merges them into one PR.
prd-taskmaster
anombyte93/prd-taskmasterTurns any goal into a validated PRD and task graph through discovery, then hands off to execution.
mind
memvid/claude-brainSearches and manages Claude's persistent memory stored in a portable .mv2 file in the project.
analyze
rsmdt/the-startupRuns a multi-agent research workflow: gathers cited claims, has skeptics refute them, writes one report.
constitution
rsmdt/the-startupUses parallel agents to discover codebase patterns and draft governance rules for user approval.
dead-ends
karanb192/claude-code-hooksLists approaches tried and abandoned in this repo, with reasons and token cost.
leaderboard
karanb192/claude-code-hooksShows which repo files cost the most context tokens across sessions, with suggested fixes.
scorecard
karanb192/claude-code-hooksShows a scorecard of which CLAUDE.md rules Claude follows and which it ignores.
Command Development
tzachbon/smart-ralphGuides writing Claude Code slash commands: frontmatter, arguments, file references and bash.
Hook Development
tzachbon/smart-ralphGuides creating Claude Code plugin hooks for events like PreToolUse and Stop, including prompt-based hooks.
learn-this
michalparkola/tapestry-skillsTakes a YouTube, article or PDF link, extracts it, and writes a Ship-Learn-Next action plan.
Plugin Settings
tzachbon/smart-ralphExplains storing per-project plugin settings and state in .claude/plugin-name.local.md files with YAML frontmatter.
Plugin Structure
tzachbon/smart-ralphLays out Claude Code plugin directories, manifest, and component discovery for commands, agents, skills and hooks.
session-log
michalparkola/tapestry-skillsSummarizes the current session into a weekly agent log of dated bullets, marking reusable outputs.
ship-learn-next
michalparkola/tapestry-skillsTurns advice, tutorials or transcripts into Ship-Learn-Next cycles with concrete shippable steps.
super-board
EricTechPro/super-boardDrives a GitHub Project board through Build, QA, Review and Done lanes via onboard, run and stop.
super-build
EricTechPro/super-boardImplements a Ready board card test-first in its own git worktree and opens a draft PR for QA.
super-collect
EricTechPro/super-boardCollects problems from project sources, verifies each one, and files the findings as board tickets.
flow
Lifecycle-Innovations-Limited/claude-opsRoutes a development task to the right lifecycle stage and can run stages as parallel agent teams.
notion-spec-to-implementation
makenotion/claude-code-notion-pluginBreaks a Notion spec into an implementation plan and tasks, then logs progress by updating those pages.
exec
umputun/cc-thingzRuns a plan file's tasks one at a time, each in an isolated subagent.
index
aerovato/operator-memoryBuilds or refreshes the Operator project index by running operator-helper and following its setup guide.
operator-index
aerovato/operator-memoryRuns the Operator Helper CLI to build or refresh a project index, upgrading and repairing it when needed.
project-init
aerovato/operator-memoryInitializes the Operator Project Brain by running operator-helper and following its setup guide.
repair
aerovato/operator-memoryDiagnoses and repairs Operator memory load failures using operator-helper memory checks.
user-init
aerovato/operator-memoryInitializes the Operator user partition by running operator-helper and following its setup guide.
make-trace
AIScientists-Dev/FlowtraceConverts a runbook or skill document into a runnable Morph trace DAG and drives it through a full run.
llm-council
gcpdev/llm-council-skillQueries ChatGPT and Gemini APIs for perspectives, then synthesizes a plan with Claude's own analysis.
collaborating-with-claude
appautomaton/latex-arxiv-SKILLLets a Codex host consult Claude Code CLI via a bridge script for reviews and unified-diff proposals.
collaborating-with-gemini
appautomaton/latex-arxiv-SKILLLets a Codex host consult Gemini CLI via a bridge script for reviews and unified-diff proposals.
agnix
agent-sh/agnixLints agent config files such as SKILL.md, CLAUDE.md, hooks and MCP configs against 757 rules.
deep-research-team
malob/nix-configCoordinates a team of researcher agents over several rounds to produce a verified, synthesized research report.
agentsop-agent-topology-selection
agentsope/SkillAlchemyDecides whether a task needs multiple agents and which topology fits, before any agent code is written.
LEAP
agentsope/SkillAlchemyRuns a staged pipeline that distills a skill from raw sources or fuses several skills into one.
Lens
agentsope/SkillAlchemyExpands a vague task into hidden dimensions, prerequisites and open questions before the agent acts.
SkillAlchemy
agentsope/SkillAlchemyGuides building an installable SKILL.md from an idea or source, or by fusing existing skills.
004-commands-installation
jabrena/plinthAsks which agent folder to use, then copies the embedded command files into it, reporting overwrites.
005-agents-installation
jabrena/plinthAsks which agent folder to use, then copies the embedded agent definitions into it, reporting overwrites.
agent-native
boshu2/agentopsLaunches independent tasks as parallel agents after checking that their write scopes do not collide.
agy-native
boshu2/agentopsRuns a supplied task in headless AGY (Antigravity) only when named, and returns the run's evidence.
claude-exec
boshu2/agentopsRuns one supplied prompt through headless Claude print mode with scoped permissions and a time limit.
codex-exec
boshu2/agentopsRuns one supplied prompt through headless Codex exec and captures its output, with no retries or batching.
antigravity
yuting0624/antigravity-for-claude-codeRoutes scaffolding, boilerplate and test generation to the Antigravity Gemini CLI for Claude to verify.
migrate-to-antigravity
yuting0624/antigravity-for-claude-codeMoves Claude Code skills, memory, MCP and permissions to Antigravity CLI, with dry-run and undo.
break-ai-slop
suyuan2022/suyuan-skillMakes the agent identify the real expert, hidden failure modes, and key constraint before a non-trivial task.
assessment
notque/vexjoy-agentRuns read-only codebase overviews, health checks, ADR reviews and decision critiques.
context-tax-audit
MiniMax-AI/mcode-stackAudits what an agent injects into each request and checks necessity, position and prompt-cache impact.
lifting-loops
MiniMax-AI/mcode-stackFinds where the user is the manual step for agents and proposes how to automate or remove it.
codex-fable5
baskduf/FableCodexSets up a Fable-style workflow in Codex: inspect first, route by cost, verify before claiming done.
claude-md-dependency-rescan
alirezarezvani/ClaudeForgeCompares manifest dependencies with the Tech Stack section of each CLAUDE.md and reports changes, read-only.
claude-md-drift-audit
alirezarezvani/ClaudeForgeChecks CLAUDE.md files against last week's git history and reports stale paths and removed dependencies.
claude-md-link-check
alirezarezvani/ClaudeForgeChecks that @imports and markdown links in CLAUDE.md files resolve to real files, read-only.
go
prekuter/dryforgeExecutes an approved dryforge plan in parallel waves, verifies with real checks, and never merges on its own.
ready
prekuter/dryforgeTurns a rough idea or spec into a grounded handoff, spec and plan, asking only the user's own decisions.
guard
Ruya-AI/cozempicStarts a cozempic background daemon that watches session size and prunes context before compaction hits.
reload
Ruya-AI/cozempicPrunes the current Claude Code session with cozempic, then opens a new terminal to resume the trimmed session.
treat
Ruya-AI/cozempicPrunes bloat from the current Claude Code session at a chosen level, showing a dry run before applying.
bro
luchasarie/bro-skillRe-explains the last reply, or pasted text after /bro, in plainer language without changing any facts.
avoid-feature-creep
waynesutton/builder-skillsKeeps code changes scoped to the request, moving extra features and refactors into a suggestions list.
anthropic-product-knowledge
syahiidkamil/Software-Engineer-AI-Agent-AtlasMakes the agent check official Anthropic docs before stating facts about Claude Code, the Claude API, or Claude.ai plans.
agent-roundtable
glebis/claude-skillsRuns a moderated debate among several coding agents that converges on one reviewed spec and task list.
agent-army
OneWave-AI/claude-skillsDeploys a two-layer parallel agent hierarchy for large refactors, migrations and audits, run by a commander.
agent-swarm-deployer
OneWave-AI/claude-skillsRuns parallel sub-agents over large data jobs, then merges results with progress tracking and error recovery.
agent-team-builder
OneWave-AI/claude-skillsInterviews you about a business process, then generates a multi-agent team config with roles and handoff rules.
organize
mcpware/cross-code-organizerLaunches the Cross-Code Organizer dashboard to view and move memories, skills, hooks and configs.
sciagent-skill-creator
jaechang-hits/SciAgent-SkillsScaffolds a new SciAgent Skills entry with valid frontmatter, registry update and validation checks.
board-of-directors
Ibrahim-3d/orchestrator-supaconductorSimulates a five-role expert board that debates a major decision and reaches a consensus.
conductor-orchestrator
Ibrahim-3d/orchestrator-supaconductorCoordinates parallel worker agents through a goal-driven evaluate loop for track execution.
ai-agent-orchestration
datadrivenconstruction/DDC_Skills_for_AI_Agents_in_ConstructionCoordinates estimator, scheduler, document, QA and safety agents under a supervisor with human checkpoints.
project-butler
JamesShi96/project-butlerInitializes and maintains project memory files such as wiki, logs, TODOs, and session handoff notes.
assumption-test
Neeeophytee/finding-unknowns-skillsHas the agent test a risky technical assumption with a small, falsifiable experiment before committing to a plan.
blindspot-pass
Neeeophytee/finding-unknowns-skillsHas the agent map blind spots, landmines and open questions in an unfamiliar area before any coding starts.
model-hierarchy
zscole/model-hierarchy-skillRoutes agent tasks to the cheapest capable model, reserving premium models for hard reasoning work.
claude-code-plugin-reference
athola/claude-night-marketExplains Claude Code plugin, skill, command, agent and hook mechanics for writing or debugging plugins.
skill-creator-ultra
marketingjuliancongdanh79-pixel/skill-generatorInterviews you about a repeated workflow, then drafts and tests a new AI agent skill from it.
dotfile-audit
harperreed/dotfilesAudits and fixes pipeline DOT files for structure, failure paths, spec coverage, and validation errors.
dotfile-brainstorm
harperreed/dotfilesTurns a vague project idea into a design doc with a pipeline configuration section for DOT generation.
dotfile-from-spec
harperreed/dotfilesGenerates and validates a headless build pipeline DOT file from a spec or design doc.
gate-control
Orkas-AI/Orkas-VideoStudioMaps user approvals and artifact state to the next allowed step in VideoStudio production gates.
genshijin-compress
InterfaceX-co-jp/genshijinCompresses natural-language memory files like CLAUDE.md into terse form to cut input tokens.
genshijin-crew
InterfaceX-co-jp/genshijinDelegates code search, edits, and review to terse-output subagents to save context.
genshijin-help
InterfaceX-co-jp/genshijinShows a one-time quick reference card for genshijin modes, skills, and default settings.
pordee
kerlos/pordeeUltra-compressed Thai and English output mode that cuts output tokens while keeping technical accuracy.
pordee-stats
kerlos/pordeeShows real session token usage and estimated savings from the Claude Code session log.
build-agent
LukeRenton/explore-claude-codeBuilds consistent Claude Code agent files with the right tools, model, permissions and memory settings.
build-skill
LukeRenton/explore-claude-codeBuilds consistent Claude Code SKILL.md files, checking for overlap with existing skills first.
core-brainstorm
LukeRenton/explore-claude-codePuts the agent in a no-code brainstorming mode that challenges assumptions and sharpens ideas before planning.
ai-brain
Ootto-AI/claude-content-skillsSaves session takeaways as Obsidian notes and recalls only relevant past notes as long-term memory.
plur-create-engrams
plur-ai/plurGuides the agent to write or refine durable memory records (engrams) from conversations, documents and decisions.
plur-memory
plur-ai/plurGives the agent persistent memory that recalls and stores learned corrections and preferences across sessions.
plur-session-end
plur-ai/plurExtracts durable corrections, conventions and decisions at session end, skipping transient or sensitive details.
generate-rules
nurettincoban/ai-prd-workflowWrites RULES.md, the coding standards AI must follow, with dependency versions checked against registries.
mine
ohad6k/emuloBuilds a personal profile from the user's own local AI coding session logs, after cost approval.
codely-plan-create-github
CodelyTV/agent-harnessTurns a GitHub issue into a plan issue with one linked sub-issue per phase, after user approval.
codely-plan-create-gitlab
CodelyTV/agent-harnessTurns a GitLab issue into a plan issue with one linked child issue per phase, after user approval.
openspec-continue-change
lis186/ccxrayCreates the next artifact of an OpenSpec change, after the user picks which change to continue.
arcgentic
Arch1eSUN/ArcgenticRuns an idea through a planned, built, self-audited, externally audited workflow in Codex.
claude-code-session-broker
Arch1eSUN/ArcgenticCoordinates fixed Planner, Developer and Auditor role sessions in Claude Code through native tools or a broker.
close-round
Arch1eSUN/ArcgenticAfter an external audit PASS, runs strict checks and marks the arcgentic round closed in its state file.
moyu
uucz/moyuKeeps the agent to minimal, requested-only changes, avoiding unrequested abstractions, rewrites and extras.
moyu-en
uucz/moyuKeeps the agent to minimal, requested-only changes and blocks unrequested abstractions and rewrites.
cross-agent-plan
prassanna-ravishankar/repowireAsks a peer running a different agent backend to draft an implementation plan you can critique and merge.
delegate
prassanna-ravishankar/repowireHands a self-contained task to a peer agent on a chosen backend and tracks it to completion via ask/ack.
repowire-patterns
prassanna-ravishankar/repowireReference for picking the right repowire primitive (ask/ack, notify, broadcast) to coordinate agents.
bootstrap
HezaoHezao/poirotRuns a short onboarding chat and writes a user profile that sets how the agent works with this user.
plan
HezaoHezao/poirotPuts the agent in plan-only mode: it writes a markdown implementation plan and changes no code.
fskill-creator
AGI-comming/functional-skill-creatorCreates, maintains, or migrates functional agent skills from a brief or a legacy skill directory.
fskill-creator-create
AGI-comming/functional-skill-creatorNormalizes a user's skill request into a create context for the fskill-creator main skill.
fskill-creator-migrate
AGI-comming/functional-skill-creatorAnalyzes a legacy skill directory into a behavior map and function split plan for migration.
creating-an-agent
ed3dai/ed3d-pluginsGuides writing Claude Code subagent files, covering description triggers, tool selection and prompt structure.
doing-a-simple-two-stage-fanout
ed3dai/ed3d-pluginsSplits a large corpus across parallel worker agents, reviews with critics, then has a summarizer write the report.
using-generic-agents
ed3dai/ed3d-pluginsGuides the agent on choosing Haiku, Sonnet or Opus for delegated work based on task complexity.
skyll
assafelovic/skyllLets the agent search the skills.sh catalog at runtime and fetch full SKILL.md content for context injection.
coordinate
agenticmail/agenticmailStarts a multi-agent task by sending one email thread that copies the team and has them take turns.
audit-agent-rules
serejaris/personal-corp-osAudits AGENTS.md rules and skill descriptions, reports what to cut or rewrite, and edits only after approval.
catchup
Avijit07x/claude-dbSummarizes where you left off in a project from saved chat memory and git into done, open and next.
cdb-scan
Avijit07x/claude-dbBuilds a code graph and written profile of the project into memory so later sessions start with context.
team-mode
oil-oil/codex-team-modeLets the main agent delegate bounded review, discovery or implementation to subagents and accept results.
antigravity-cli
scarletkc/agentsDelegate bounded review or coding tasks to the Antigravity CLI and continue them in one conversation.
ask-to-plan
scarletkc/agentsGuide a rough idea to an actionable plan by asking one native multiple-choice question at a time.
codex-cli
scarletkc/agentsGuide when to hand work to Codex CLI for second-model review, with sandbox, model, and effort choices.
desk-operating-model
galleonlabs/hypergrok-trading-deskDefines roles, seats, shared workspace, approval model and handoff format for a team of Grok Bots.
devin-cli
scarletkc/agentsDelegate bounded coding tasks to Devin CLI over ACP, with single-use permissions and resumable sessions.
grok-cli
scarletkc/agentsDelegate bounded tasks to Grok Build over ACP, with background progress checks and follow-up prompts.
code-standards
tamdogood/builder-essential-skillsWalks any code change through orient, baseline, smallest fix, tests, verification and self-review.
compact-adviser
kunchenguid/compact-adviserReads and changes compact-adviser settings for Codex: hint mode, token thresholds, request logging and API key.
deep-project
piercelamb/deep-projectSplits vague project requirements into numbered spec folders and a manifest for later /deep-plan planning.
court
alexyc9381/court-skillPuts an idea on trial with a prosecutor, defense and 12 juror Claudes, then gives a verdict and flip conditions.
agent-instructions
murphytrueman/design-system-opsWrites an AGENTS.md that tells coding agents where design system parts live, its rules and what to avoid.
agentic-sop
s0912758806p/agentic-sop-to-workTurns a manual SOP into a gated agent workflow with a topology, typed handoffs and a human approval stop.
agentic-workflow-audit
s0912758806p/agentic-sop-to-workRead-only audit of an agent pipeline that checks task splits, data contracts, shared state and retry loops with evidence.
six-rung-ladder
s0912758806p/agentic-sop-to-workApplies a staged filter before adding code, dependencies or abstractions, starting with whether it is needed at all.
sop-flow
s0912758806p/agentic-sop-to-workRuns the agentic-sop-kit workflow on a project or the bundled demo, then summarizes the DRAFT report and manifest.
procoder
azrtydxb/procoderRuns the procoder commit gate and drives its spec, plan and todo chain before the agent calls work finished.
skill-align
techygarg/latticeAudits and fixes Lattice docs and templates so they match the current skill inventory.
skill-enhance
techygarg/latticeUpgrades an existing Lattice skill through rewrite, behavior checks, QA and independent verification.
skill-forge
techygarg/latticeCreates new Lattice atom, molecule or refiner skill files that follow the framework conventions.
skill-review
techygarg/latticeAudits a Lattice skill through three review personas and reports practical behavioral gaps with fixes.
skill-tighten
techygarg/latticeAudits a SKILL.md for soft wording and missing enforcement, then rewrites it into firm imperatives.
clarify-intent-and-establish-shared-understanding
Inference1/clarify-intent-and-establish-shared-understandingAsks layered, adaptive questions to test a plan's assumptions, risks and goals until intent is shared.
recall-context
davegoldblatt/total-recallShows which memory files are loaded in the session and which exist on disk.
recall-forget
davegoldblatt/total-recallMarks matching memory entries as superseded across memory tiers without deleting them, keeping history.
recall-init
davegoldblatt/total-recallCreates the Total Recall memory folders, working memory file and schema, skipping any that exist.
recall-init-ids
davegoldblatt/total-recallAdds durable ^tr IDs to untagged memory list entries so maintenance commands can track them.
recall-log
davegoldblatt/total-recallAppends a timestamped raw note to today's daily memory log without any write gate.
skill-conductor
smixs/skill-conductorCreates, tests, improves and packages agent skills through a draft, eval and review loop.
trace-mcp-codemod
nikolai-vysotskyi/trace-mcpMakes the agent apply repeated mechanical edits through the apply_codemod tool instead of many Edit calls.
decompose-feature
FrkAk/piyazBreaks a new feature into 5 to 20 linked tasks inside an existing Piyaz project without changing its status.
clarify
mag123c/toktrackClarifies requirements adaptively before coding, escalating from quick Q&A to deep exploration and a plan.
agentctl-setup
sd0xdev/sd0x-harnessInstalls and sets up the agentctl mod for task scope limits, pre-run refusals and stale test evidence.
skill-generalizer
hqhq1025/skill-optimizerTurns a private local skill into a public-ready one by removing paths, credentials, and personal context.
skill-miner
hqhq1025/skill-optimizerMines past agent sessions and memories for repeated workflows and drafts candidate new skills.
auditing-progress
prime-radiant-inc/iterative-developmentAudits each iteration's behavior evidence in three tiers using paired adversarial subagent reviews.
acc
automatis-tools/agents-can-communicateCoordinates messages and handoffs between independent local AI sessions through ACC's own store.
readme-skill
study8677/Readme.skillBuilds a shareable README profiling your AI coding tool usage from local, read-only history.
agent-skills-creator
mblode/agent-skillsCreates and audits portable Agent Skills, with validation, routing checks, and keep or cut decisions.
skillhone-auto-optimization
Tencent/SkillHoneRoutes a reproducible defect found during skill use into a tracked Issue and a tested pull request.
skillhone-benchmark-optimization
Tencent/SkillHoneRuns a frozen benchmark eval loop to pick the next Skill failure to fix and record results in a PR.
atdd-team
swingerman/engineerOrchestrates a six-phase ATDD feature workflow, running each phase with a fresh agent.
token-saver-config
ppgranger/token-saverConfigures and diagnoses token-saver compression settings, routing rules and savings statistics.
codex-model-routing-team
zjp1997720/codex-model-routing-teamSplits parallel work into background Codex tasks with chosen models and effort, coordinated by a lead agent.
agent-routing
oaustegard/claude-skillsPicks each subagent's model, effort, and cascade by measured cost per completed task.
check-skill
UiPath/coder_evalBuilds and runs an activation test suite to check whether a Claude Code skill triggers as intended.
bulletproof
artemiimillier/bulletproofGuides the agent through an adaptive 12-stage workflow for non-trivial coding tasks, from research to deploy.
research-mode
assafkip/research-modeToggles a mode requiring citations, source grounding, and admitting unknowns until the user turns it off.
recallloom
Frappucc1no/recall-loomKeeps project state in durable files so the agent can resume long-running work across sessions.
roadmap
Ovid/paadReads the feature roadmap, picks the next unplanned phase, brainstorms it and updates the roadmap file.
lintlang
hermes-labs-ai/lintlangLints agent instructions and tool definitions offline for ambiguity, missing stop rules and mixed formats.
audit-agent-settings
vinta/hal-9000Audits Claude Code or Codex settings against the latest official docs and proposes tailored changes.
antigravity-agents
markfulton/claude-antigravity-agentsDelegates coding, review and research jobs to Antigravity agy sub-agents that run in parallel.
agent-wiki-compare-outcomes
AgentToolkit/altk-evolveDerives evidence-backed guidelines by contrasting successful and failed agent trajectories on the same task.
agent-wiki-consolidate-guidelines
AgentToolkit/altk-evolveGroups near-duplicate atomic guidelines into themed cluster pages while keeping the original pages.
faq-shortcuts
kishormorol/cli-faq-shortcutsMines your past agent session history for repeated asks and turns them into project slash-command shortcuts.
usage-governor
0xhimanshu/governorMakes the agent answer concisely and avoid token-wasting habits to stretch Claude Code usage limits.
wonder-pill
ara-mkr/Wonder-PillTurns open questions into assumption-testing what-if provocations and a mind map, not ready-made answers.
wonder-pill-universal
ara-mkr/Wonder-PillPortable version that turns open questions into assumption-testing what-ifs and a mind map on any assistant.
alive-health
alivecontext/aliveChecks ALIVE walnut projects against their update rhythm and flags stalled ones on a daily schedule.
alive-inbox
alivecontext/aliveScans the ALIVE inbox folder for unrouted files and suggests which walnut each belongs to.
brain-jam-plan
povvo/claudikins-kernelRuns iterative, human-in-the-loop planning with one question at a time and explicit trade-off checkpoints.
cco
egorfedorov/claude-context-optimizerShows session cost and context use in Claude Code, with commands to audit, pack and trim tokens.
habit-words
joetawil7/first-passFinds the habitual words a user types that weaken answers, and maps each to the checks it should trigger.
janitor-discover
khendzel/skills-janitorSearches GitHub for agent skills and checks a skill's overlap and risk before you install it.
janitor-fix
khendzel/skills-janitorPreviews fixes for broken skill frontmatter, applies them on request, and can prune broken symlinks.
ui-agent-patterns
Ormus-Solutions/LibreUIUX-Claude-CodePatterns for splitting UI work across one synthesis agent or specialist agents, with clean handoffs.
ax-extract-workflow
Necmttn/axReconstructs the ordered skill steps behind a past commit, date or feature from ax session history.
odai
orziz/odaiGoverns general tasks: clarify the real goal, verify facts, act within authority, and check results.
odai-orchestration
orziz/odaiCoordinates optional role-based sub-work for planning, review and parallel tasks, with model routing setup.
advocatus-diaboli
Habitat-Thinking/ai-literacy-superpowersChallenges a spec with evidence-backed objections across six categories before plan approval.
ai-literacy-assessment
Habitat-Thinking/ai-literacy-superpowersScores a team's AI collaboration literacy from repo evidence and questions, then writes a dated report.
agent-briefing
NoobyGains/godmodeTeaches the agent to write complete briefs for subagents and require checkable report files as handoffs.
winnow
GhalebDweikat/winnowExplains how to read winnow stubs in tool output and restore hidden blocks with winnow_recall.
prompt-sensei
chengzhongwei/Prompt-senseiCoaches prompt quality in AI coding sessions, scoring and improving prompts and reporting habits locally.
goal-setter
gotalab/goal-setter-skillTurns a rough request into a compact verifiable Goal with evidence, boundaries and stop conditions for Codex.
sync
Umair/stalemapWrites a block in AGENTS.md or CLAUDE.md listing npm packages changed after the model's training cutoff.
upgrade
Umair/stalemapMakes the agent check npm package API changes via stalemap before writing or upgrading code that uses them.
dsh-delegate-runtime
cpj-dev/dsh-plugin-ccDefines how commands call the DeepSeek Harness bridge: subcommands, flags, environment and failure handling.
dsh-run-output
cpj-dev/dsh-plugin-ccShows the agent how to present DeepSeek Harness bridge results verbatim, keeping run IDs and footers.
deep-plan
piercelamb/deep-planRuns a staged planning flow of research, interview, external LLM review, and a TDD implementation plan.
obsidian-knowledge-brain
Tubo2333/obsidian-knowledge-brainCaptures errors and decisions from each agent session into a project-local searchable knowledge base.
dash
tomtomsatnav/agent-orchestration-standardShows a snapshot of the AOS agent dashboard with agents, status, tokens and file conflicts.
Nothing matches. Try a shorter word.