GITHUB AGENT SKILLS / 技能收录
给你的 AI 助手
装上专业技能
我们从 GitHub 持续收录、审核并分类优质的 Agent Skills,按职业与使用场景整理,复制一行命令即可安装到 Claude Code、Cursor 等智能体。
按职业浏览 Agent Skills
为不同岗位挑选的高频技能
按使用场景浏览 Agent Skills
从你要完成的任务出发
按创作者浏览 Agent Skills
按 GitHub Stars 排序的优秀创作者
Microsoft
@microsoftopenclaw
@openclawAnthropic
@anthropicsOpenAI
@openaiThe Apache Software Foundation
@apacheGitHub
@githubAutoGPT
@Significant-GravitasGarry Tan
@garrytan全部 Skills 482556
flow-mapping
Turn current-state data-flow evidence into Mermaid diagrams that compile and state payload, cadence and reliability, gated by a validation script, at .gener8v/flows/ with an explicit Unknowns list. Use for data-flow, system-map or integration-map deliverables, or to check existing diagrams before they are published, such as 'map how claims data moves between systems' or 'draw the integration map'. Current state only: not for future-state architecture proposals (technical-design).
dependencies
Map dependencies between capability areas, external systems and shared resources into .gener8v/dependencies/dependency-map.md, with phased sequencing, parallelization opportunities and the critical path. Use when a PRD has more than one capability area and implementation order must be decided, such as 'what order should we build these in', 'what is the critical path' or 'what can run in parallel'. Not for the constraints on one area (constraints) or ordering tickets inside one area (ticket-breakdown).
delivery
Implement one ticket: reconcile its assumptions against the real codebase, get an implementation plan approved, write the code with @spec annotations, and keep the delivery record current from plan approval onward. The only skill that changes source code. Use when a ticket's dependencies are delivered and it is ready to build, such as 'implement TICKET-004' or 'deliver the next ticket'. Not for writing tickets (ticket-breakdown) or reviewing delivered code (code-review).
defect-sweep
Fresh-context sweep of an existing subsystem's perimeter for defect classes nobody is looking for (inherited defaults, sibling writers, incomplete gates, fail-open error paths, silent truncation, stale assertions, identity confusion, time-based inference, unasked authorization), with proof per finding, at .gener8v/sweeps/. Use after a burst of changes, before a launch, or when a found defect may have siblings, such as 'are there more bugs like this one' or 'sweep the billing module'. Not for one delivery (code-review), decay over time (rot-watch) or pipeline documents (audit).
constraints
Surface technical, compliance, integration and operational constraints (TC/CC/IC/OC-XXX) from the PRD or one specification, each with rationale, affected requirement IDs, interactions and risk flags. Use before technical design or ticket breakdown when the system has regulatory, infrastructure or integration boundaries, such as 'what compliance rules apply to this' or 'what limits does the legacy system put on us'. Not for choosing an architecture (technical-design) or ordering work across areas (dependencies).
code-review
Review one delivered ticket against the pipeline: acceptance criteria, requirement coverage, constraints, architecture decisions and @spec annotations, producing traceability tables, findings and a verdict. Use after a delivery, in parallel with quality-review and security-review, when asked to 'review TICKET-003', 'check the delivery against the spec' or 'did we build what the ticket asked for'. Not for readability or tests on their own (quality-review), vulnerabilities (security-review), or code with no ticket behind it (quality-review on the files).
brownfield
Map an existing codebase into the gener8v pipeline bottom-up: system context, one specification per capability area with requirement IDs, a PRD synthesized from the code, and @spec annotations in source. Use when onboarding existing code that has no .gener8v/prd.md, such as 'bring this repo into the pipeline' or 'retrofit specs onto this code'. Not for a product with no code yet (planning) or a project that already has a PRD (orchestrate).
audit
Audit pipeline artifacts for gaps, inconsistencies, missing coverage and staleness: one document, cross-stage traceability across .gener8v/, or a ticket or externally-authored plan reconciled against the real codebase (Go / Blocked), then resolve the findings interactively. Use when asked to 'check the spec for gaps', 'sanity-check this plan before we build it' or 'does this ticket still match the code', before committing to a stage, or at milestones. Audits documents and plans, not code: for delivered code use code-review, quality-review or security-review.
architecture-review
Adversarial, code-grounded review of an existing system's architecture: do the ADRs match the code, is the design right for a system of this nature, what are its inherent limitations and what concretely would have been better, every alternative proven against the code. Use when asked 'is our architecture sound' or 'what should we change about the design', for technical due diligence, before a major architectural commitment, or at milestone health checks. Not for one delivered ticket (code-review), decay across deliveries (rot-watch) or defect hunting (defect-sweep).
notes
Read a note back with the notes_echo tool.
workflow-builder
Turn a repeatable business process described in conversation into a named, validated ArcFlow workflow draft.
update-tool
Update an existing @tool source file in workspace/capabilities/, bumping its semver per major/minor/patch.
update-skill
Update an existing skill's SKILL.md body in workspace/capabilities/skills/, bumping its frontmatter semver.
create-tool
Author a new @tool-decorated Python file in the agent workspace, validate it, and register it via reload().
create-skill
Scaffold a new skill folder under workspace/capabilities/skills/ with frontmatter and the seven required sections.
sqlite
Use the SQLite connector to read the operator's semantic layer and make bounded typed reads of an approved database file. Trigger for questions about data in a connected SQLite database; skip for writes, schema changes, or raw SQL.
slack
How to read and search a connected Slack workspace — listing the channels and DMs you belong to, reading a channel's or thread's recent messages, searching your messages, and (when allow-listed) posting a message. TRIGGER when a request asks what was said in Slack, to find a Slack message or thread, to summarize a channel, or to post to Slack. SKIP when no Slack connection is granted, or for a channel or DM you are not a member of.
s3
Use the S3/MinIO connector to enumerate an approved bucket prefix and read one bounded object. Trigger for connected object storage; skip for uploads, deletes, bucket-policy changes, or arbitrary prefixes.