repo-development
Guides CrewAI repository contributors through monorepo layout, workspace package metadata, focused pytest and lint checks, docs versioning, release-freeze hazards, and safe native verification selection.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- repo-development
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Guides CrewAI repository contributors through monorepo layout, workspace package metadata, focused pytest and lint checks, docs versioning, release-freeze hazards, and safe native verification selection.
- disable-model-invocation
- true
02 资源文件 5
- 参考文档references/contributor-guidance.md6.1 KB
- 参考文档references/docs-maintenance.md6.1 KB
- 参考文档references/testing.md7.2 KB
- 参考文档references/troubleshooting.md6.5 KB
- 脚本scripts/select_native_tests.py14.6 KB
03 SKILL.md 正文
Repo Development
Use this sub-skill when work is about contributing to the CrewAI monorepo itself: choosing focused checks, understanding workspace packages, editing docs safely, avoiding release/docs snapshot hazards, or selecting safe native verification targets after a code or docs change.
Read First
- Contributor guidance for repository layout, workspace packages, package metadata, contributor workflow, and cross-package dependency pins.
- Docs maintenance for Edge docs, frozen snapshots,
docs/docs.json, image rules, docs validation, and release-freeze no-run guidance. - Testing for root pytest, ruff, mypy, devtools test configuration, focused command selection, and safe native verification principles.
- Troubleshooting for frozen-doc edits, image deletions, broad tests, import-mode confusion, workspace pin drift, and accidental release-freeze mutations.
- Native test selector to map changed paths to suggested focused checks without running tests, release scripts, LLMs, credentials, or network calls.
Boundaries
- Stay here for monorepo contribution mechanics, docs versioning, package workspace metadata, test/lint/type-check selection, docs freeze hazards, and maintainer-only devtools guidance.
- Use ../../SKILL.md when a request needs root CrewAI scenario routing before deciding whether it is repository-maintenance work or normal CrewAI usage.
- Use ../cli-and-projects/SKILL.md for
crewaiCLI behavior, project scaffolding, command semantics, and project template troubleshooting beyond repository test selection. - Use ../core-runtime/SKILL.md for
Agent,Task,Crew, processes, guardrails, callbacks, and runtime semantics. - Use ../flows-and-events/SKILL.md for
Flow, decorators, routers, persistence, event listeners, and hook adjacency. - Use ../tools-and-mcp/SKILL.md, ../memory-knowledge-and-rag/SKILL.md, ../llm-and-providers/SKILL.md, ../files-and-multimodal/SKILL.md, or ../observability-and-hooks/SKILL.md for normal CrewAI package usage and capability-specific API details.
Safe Operating Defaults
- Prefer the smallest relevant check first: a focused pytest file,
crewai --help, parser/help output from bundled scripts,ruff check --no-fix, or docs link checks. - Do not edit
docs/v*/snapshots during normal development; updatedocs/edge/<lang>/...instead. - Do not delete or rename existing
docs/images/assets; add a replacement asset and point Edge docs at the new file. - Treat docs freeze and release automation as no-run unless the task is explicitly a release-cut context with the expected
[docs-freeze]workflow. - Use the bundled selector before broad native verification when a task changes both code and docs.
Usability Targets
- After a CLI command plus docs change, select safe focused checks covering CLI command loading, changed command tests, Edge docs validation, and any docs/devtools impacts without running LLM-backed project execution.
- When updating docs, keep edits in Edge, preserve frozen snapshots, avoid image deletion/renames, update navigation only when needed, and choose validation commands that do not mutate release state.
04 同仓库更多 Skills
passnet-triton-opt
Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.
passnet-solve
END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.
passnet-skill
PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.
passnet-pattern-fusion
Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.
passnet-orchestrate
Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.
passnet-feedback
Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).