Index of every page on this site: llms.txt
All skills
scode avatar

/scode-commit-msg-reviewer

@03a9f7f
Run it yourself

Use only when the user explicitly invokes scode-commit-msg-reviewer by name, asks for commit messages to be checked by a fresh-context reader, or their standing instructions say to always use this skill. Do not trigger merely because a commit or PR is being created. Puts commit messages, PR titles/descriptions, and equivalent VCS prose in front of a cold reader who has not seen the change, and rewrites until that reader understands them; after the first signal, applies to all commit-message-like content for the rest of the conversation. Also use when the user says "scode-commit-msg-reviewer feedback: ..." to record feedback about how this skill performed.

  • 4 files
  • 21.9 KB
  • Updated 3 months ago
  • GitHub
Use this Skill: https://skilld.dev/gh/scode/dotfiles/scode-commit-msg-reviewer

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

eval.md

≈535 tokens on demand. Your agent reads this file only when SKILL.md points to it.

Evaluating changes to this skill

NOTE: This is not part of the skill. Nothing in SKILL.md or reviewer.md refers to it, and an agent running the skill should never read it. It exists so that a change to the reader's charter or the loop can be checked against a known set of commits without hunting for them again.

Commits

PR commit title
#300 93b6697 feat: make modernize ask whether a repo is public or private
#301 bddc7ba feat: prefer tailscale ssh for tailnet workers in scode-ssh-delegate
#302 2d30f74 feat: add a one-shot fast path merge to jjstack
#303 63630ed feat: add muse as an opt-in delegation family to scode-galaxy-brain
#304 1c7f1b4 fix: stop telling muse delegates to run with write policy disabled
#305 51ee3e1 feat: warn up front that scode-galaxy-brain skips permission checks

Procedure

One drafting run per commit per version of the skill. Each run is a fresh agent given the commit's diff as a file (git show output with the message stripped) and not the hash, told not to consult git history, and told to write the commit message as if it had just authored the change, using scode-commit-msg-reviewer by name and running its loop with its own fresh reader subagents until the loop says done or reaches its cap. Have it write the final message, the number of readers spawned, and every round's candidate and reader report to a file.

~/.claude/skills and ~/.codex/skills symlink into this repository's working copy, so a version of the skill is put under test by placing its SKILL.md and reviewer.md in the working copy. Runs for one version can go concurrently; versions go one after another; do not run jj while an uncommitted version is swapped in, and restore the committed state afterward.

Compare the final messages against the real ones and against each other, and read the round logs. Change one thing per version.

Source: SKILL.md on GitHub

No rule matched.

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

No third-party reports yet.

Signed by skilld at 03a9f7f. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub yesterday.

Activeupdated 3 months ago

README badge

README badge for scode/dotfiles/scode-commit-msg-reviewer