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
@microsoft
openclaw
@openclaw
OpenAI
@openai
The Apache Software Foundation
@apache
GitHub
@github
Anthropic
@anthropics
AutoGPT
@Significant-Gravitas
Garry Tan
@garrytan全部 Skills 196921
codex-triage
Use when you need a fast classification of next steps before deeper work.
codex-session-reunion
Use when Codex sessions or rollout JSONL files are spread across Windows, WSL, backups, or VS Code mirrors and need to be inventoried, deduped, and promoted into one canonical store safely.
codex-runtime-recovery
Use when Codex state is split across Windows, WSL, or VS Code, sessions disappear, state_5.sqlite files churn, or CODEX_HOME and launch routing drift need diagnosis before destructive repair.
codex-review
Use when you want one Codex critic to review a plan or change before implementation.
codex-orchestrate
Route a task to Codex on the MCP rail by intent - one skill, all modes. Trigger phrases - 'brainstorm/discuss/coordinate/review/audit/curate/triage with Codex', 'get a Codex second opinion', 'have Codex split roles and combine', 'pick the best option with Codex', 'Codex risk/gap audit', 'Codex governance call'. The rail is the official codex mcp-server - mcp__codex__codex to start a thread, mcp__codex__codex-reply to continue by threadId. There is no per-mode tool - the mode is framed in the prompt, and multi-peer modes are parallel calls the caller synthesizes. Replaces the per-mode wrappers codex-audit, codex-brainstorm, codex-coordinate, codex-curate, codex-consult, codex-discuss, codex-fix, codex-governance, codex-review, and codex-triage. For per-call scoping (model, sandbox, MCP allowlist, --output-schema) use codex-invoke instead.
codex-mixed-workflow-doctrine
How to put OpenAI Codex to work from Claude on this Windows box, and how to run mixed Claude+Codex workflows correctly. Covers the four rails (native codex MCP, detached codex-exec fan-out, the mingle peer channel, a Claude Workflow), the canonical Windows detached fan-out recipe with every hard-won flag, the terminating-unit spec template, the standard mixed-model topologies, the full failure catalogue with exact signatures + fixes, and the two-signal verification discipline. Use when fanning out Codex, designing a Claude+Codex handoff, or diagnosing a stalled/hung/false-green fan-out.
codex-invoke
The ONE canonical entry point for invoking OpenAI Codex with per-call scoping. Use whenever you run `codex exec` (or drive the `mcp__codex__codex` / `codex-reply` rail) with a specific model, sandbox/approval policy, MCP allowlist or seal, skills allowlist, or structured `--output-schema` output. NEVER hand-roll `codex exec` argv. NOT for MCP-rail routing (codex-orchestrate).
codex-governance
Use when you need a final governance-oriented decision across options and risks.
codex-fleet-github
Use when dispatching, sweeping, monitoring, or planning autonomous Codex-on-GitHub runs across the private fleet via the codex-fleet-github MCP server. Drives the Wave-1 workflow_dispatch runner (codex-runner.yml) with propose-only draft PRs and a fail-closed auth model. Covers fleet_dispatch, fleet_sweep, fleet_status, fleet_orchestrate. NOT for local Codex CLI consultation (use codex-consult).
codex-fix
Use when you want Codex to coordinate a fix path across multiple reasoning roles.
codex-discuss
Use when you want two Codex peers to debate tradeoffs before synthesis.
codex-databricks-fanout
Use when fanning out Codex CLI agents onto Databricks - running many parallel Codex/LLM work units on job clusters, scaling past a local RAM ceiling, or deciding between Databricks / local / Modal / RunPod for a large agent fan-out. Carries the measured constraints from a 608-unit production run, a 48-unit auth-research re-prove, and CRCP ops so a session does not re-derive them wrong.
codex-curate
Use when you want a final best-option selector, optionally with PromptBridge talk-only curation.
codex-coordinate
Use when you want Codex to split roles and consolidate the outcome.
codex-consult
RETIRED 2026-08-09 - redirect tombstone. The PromptBridge/Agentify Codex-side consultation lane this skill drove was deleted from the repo; its codex_local_consult / codex_local_curate / codex_local_review and codex_orchestrator_* tools do not exist and cannot be mounted. Kept only so in-flight invocations of the old name resolve. For a Codex consult / curate / review pass use codex-orchestrate (MCP rail) or codex-invoke (exec rail, per-call scoping); both ride mcp__codex__codex / mcp__codex__codex-reply.
codex-brainstorm
Use when you want two Codex peers to generate ideas before synthesis.
codex-audit
Use when you want an audit-style orchestration pass over risk, gaps, and verification needs.
code-security
Security guidelines for writing secure code. Use when writing code, reviewing code for vulnerabilities, or asked to 'check for SQL injection' or 'review security'. IMPORTANT: Always consult this when writing or reviewing code that handles user input, authentication, file operations, database queries, network requests, cryptography, or infra config - even if security is not mentioned.