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
requirements-interview
The Requirements stage of the plan-gate — fills plan.md's Requirements section via structured Q&A where the agent proposes a cited best-guess per question and the supervisor reviews. Reached through plan-gate (which owns the brainstorm→plan entry), not as an independent trigger.
request-supervisor
Decision protocol for when to act autonomously vs escalate to supervisor. Use before any communication with the user or parent-agent, or whenever uncertain whether to act autonomously or escalate.
rack-advanced
Advanced rack beyond day-to-day lifecycle: authoring a new custom backlog, and cross-project work. Use when defining a backlog beyond the built-in tasks/HYP/PROP, or registering and searching backlogs across several projects.
predictive-accounting
For shrink/refactor/optimize/cleanup tasks — present baseline + predicted delta + dependency cost BEFORE implementation, not after
positive-control
Distinguish a failed subject from a broken verification path. Use before relying on search or grep output, logs, composed prompts, stored artifacts, or before/after measurements to support a verdict.
plan-gate
The single entry point for the brainstorm→plan quality gate — a table of contents mapping each required plan section to the skill that fills it. Use when a task enters brainstorm→plan or when drafting/revising plan.md, to route each section to its owning skill before requesting plan→execute.
plan-discipline
Put a plan in its one canonical home, the tracker `plan.md`, never `.claude/plans`. Use on entering plan mode, on any plan draft including ExitPlanMode output, or when a task enters brainstorm→plan.
judge-role-protocol
HITL dialogue + CLI ops contract for role-judge — file-fix tasks instead of editing role/skill/rule sources
judge-protocol
HITL protocol for the judge role (Phase 2 of the two-phase judge split) — discuss the Phase 1 draft with the supervisor, ack and execute proposed mutations via CLI, write the final report.
judge-auditor-protocol
Read-only audit protocol for the judge-auditor role (Phase 1 of the two-phase judge split) — produces a draft report with proposed verdicts and mutations for the HITL judge to consume.
interface-design
Shape a public contract so a reader with none of your context can use it. Use on any CRUD to a public interface — a field, method, type or enum member added, retyped, renamed or removed — and when reviewing a diff that touches one.
hatrack
Backlog lifecycle on the `rack` CLI — create / ls / context / transition / plan-extract, with field edits via `--set` and hypotheses/proposals on the `rack hyp` / `rack proposal` groups. Use for any task lifecycle transition, backlog read, field edit, document, link, hypothesis, or proposal.
git-mastery
Advanced git operations covering branches, conventional commits, worktrees, and rebasing. Use for any git operation beyond basic add/commit/push, for branch management, rebasing, and conflict resolution, or when setting up commit conventions for a project.
devils-advocate
The Approach & counter stage of the plan-gate — a skeptic value-counter that steelmans the value, then challenges whether the work is worth doing (needed? missed anything? another way?) before scoping. Reached through plan-gate, not as an independent trigger.
design-minimalism
The Scope & Out-of-scope stage of the plan-gate — every primitive (class, abstraction, option, new step) in plan.md must be justified by a concrete current-epic use case; speculative additions go to Out of scope. Reached through plan-gate, not as an independent trigger.
context-reset
Clean context reset protocol covering commit work, write handoff, update task card, and inform supervisor. Use when the context window is approaching its limit (system signals compression), when switching to a fundamentally different task domain, or when explicitly requested by the supervisor.
context-handoff
Summarize critical context (decisions, forks, pitfalls) into a handoff file for the next agent. Use before a context reset (invoked by context-reset), when explicitly asked to save current context, or before delegating work to a sub-agent that needs background.
command-lifetime
Diagnose a shell command refused or flagged for having no bound on its lifetime — an unbounded while/until loop, an unbounded background launch, or a long-running command with no timeout.