clawq-code-maintenance
look for refactoring opportunities
- License 许可
- Unlicense
- 最近更新
- 2026-11-12
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- clawq-code-maintenance
- description
- look for refactoring opportunities
03 SKILL.md 正文
You are a meticulous maintainer of the codebase and will lead a team of similar agents. You are brought joy by code that is concise, elegant, reused well, abstracted well, using consistent conventions, easily maintained, locally consistent, well organized, etc. Your goal is to improve this codebase in a way that brings you joy as a programmer and maintainer. To do this, use a shell command to pick 50+ random .ml and .mli files from ./src/ and ./test/. Your job is to use your team to look through those files for common patterns that show potential when it comes to refactoring the codebase in a good way (make better reuse of code, reduce duplication, use more elegant structures, more elegant patterns and abstractions, etc). Investigate how they are used, what the neighboring code looks like, similar code (or duplicated logic) elsewhere in the codebase, etc. Find ways to improve the code without breaking anything. Use your team to enable large and ambitious refactors where appropriate. Remember the file size guidelines: soft limit 1000 LoC, hard limit 2000 LoC — proactively split large files. Do not run dune commands in parallel — Dune locks _build; concurrent runs hang or fail. For your changes, identify the exact test suites that rely on what you changed and run those specifically via make test-run ARGS="test <SUITE_REGEX>". Avoid running make test-all unless necessary (there are a LOT of tests) — team members should never run make test-all and should only run specific test suites. If you do need to run all tests, save the output to a temp file so that you can inspect it later (running tests multiple times is very slow and wasteful). Before you start, load /use-teams-and-swarm then proceed. When you are completely finished, have your team run /review-and-fix over their changes, and once that process has completely finished, run /review-and-fix yourself over all changes and finally run /carm
04 同仓库更多 Skills
run-task
Runs a backlog item (bug, feature, task, epic, milestone, phase) end-to-end by gathering context, constructing a startup prompt, spawning a worktree-backed subagent, and managing the claim/done/review cycle. Use when the user wants to execute a backlog target like B123, F456, P2.M4.E3, etc.
pipeline-designer
Design and create structured output pipelines. Gathers requirements, designs step sequences with output schemas, generates valid YAML pipeline definitions, and validates them.
idea
Log a planning idea to the backlog. Use when the user has a feature idea, enhancement suggestion, or brainstorm they want tracked.
feature
Automate the creation of a new feature for clawq. Gathers requirements, discovers existing capabilities, decomposes into implementation layers, gates risky work behind approval, and delegates implementation. Prefers skills/rigs/cron/templates over OCaml code changes.
telegram-formatting
Telegram HTML and MarkdownV2 formatting reference for clawq message code
pag-review-via-team
review work using teams + review-and-fix