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
skill-forge
Distill a hard problem you just solved into a reusable skill draft - proposal-only, behind a three-condition threshold. Use after finishing a non-trivial task whose procedure would otherwise die with the session. Trigger with /skill-forge.
session-history
Search explicitly selected local Codex or Claude JSONL session logs to recover past work with source file and line evidence. Use when the user asks about earlier coding sessions or decisions.
review
Review current changes or specified files - ruff lint for Python files, an adversarial review-code pass, and a security scan only when security-sensitive files changed.
plan-task
Structured task planning into a plan file. Use for any new feature, bug fix, or project that needs a task breakdown before implementation.
multi-perspective-review
Multi-perspective review of a plan or design document. Up to ten viewpoints deliberate in 3 rounds; the main session orchestrates, with a sequential path for harnesses without subagents.
harness-export
Export the kit's host-neutral rules to every harness entrypoint file (AGENTS.md, GEMINI.md) so non-Claude-Code harnesses (Codex, OpenCode, Copilot, Gemini CLI, Pi, Hermes) follow the same discipline. Use when a repo is worked by more than one agent harness, or after changing rules/. Trigger with /harness-export.
debug
Investigate a failure whose cause is not evident from the error itself - flaky or intermittent failures, failures that only happen in CI or one environment, or errors with several plausible causes across files - by reproducing, isolating, fixing and verifying. Skip it when the message already shows the fix (a missing variable, a typo, a wrong import); just fix that directly.
cross-engine-review
Reach an evidence-backed decision between sessions running on different agent engines. Use when a call is high-stakes enough that one engine's blind spots are a real risk, and a second session with different tool access can bring evidence the first cannot.
control-loop
Multi-session control discipline for delegating work to worker sessions — investigate, decide, dispatch, verify, and merge. Use when a session is coordinating separate worker sessions or worktrees and needs to know when to delegate versus decide, and how to verify before merging.
code-intel
Use AST search, replacement previews, and language-server diagnostics or navigation in trusted workspaces; servers may execute project code. Choose AST for code structure, LSP for semantic questions, and text search for literal strings.
child-session
Discipline a worker session loads at dispatch time — verify the brief's premises before implementing, work only on its own branch, and report in the fixed completion format. Use when a session has been dispatched as a worker (child) with a brief from a parent control session.
brainstorming
Design and spec a Large new feature or system change before planning - explores requirements, proposes approaches, gets design approval, then chains to plan-task. Skip for bug fixes and Small/Medium work; go straight to plan-task or just do it.
auto-dev
Automated development pipeline. Runs a completed plan file through Development and Validation phases end-to-end.
agent-creator
Create a Claude Code sub-agent definition for this project or user. Use when the user asks to create a new custom agent or a specialized sub-agent for a recurring task.
zenifra
Use when the user wants to manage Zenifra projects, profiles, organizations, deployments, builds, authentication, or supported Git providers from Codex.
hyper-review
暂无简介
hyper-debate
暂无简介
hyper-cleanup
Use this skill when someone asks to clean up, delete, or consolidate code, tests, comments, or dead weight in a repository that already works, and the cleanup must not change behavior. It names each in-scope unit as a slop category or a keeper, walks a delete-reuse-platform-simplify ladder, locks behavior with a runnable baseline before the first edit, proves each test deletion or consolidation with a source mutation that is restored byte for byte, and writes a report under `.hyper/cleanup/`. Do not use for authoring new behavior, refactoring a public API, migrating dependencies, or formatting a whole repository.