agentleFS
Nicolas-RigaudyNicolas-Rigaudy/cr-0w/skills/mr/SKILL.md

mr

Skill1 starsChanged 6 days agoMITView on GitHub

Open a merge request (GitLab) or pull request (GitHub) the user's way — repo template, title validated first, user assigned, reviewer asked. Use whenever you are about to create an MR/PR, from the hub or from a repo session, or when the user says "open the MR", "create a PR", "mr".

  • Commits and pushes

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 2 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Distinctive:

    Every MR/PR you open for the user follows this flow.

    line 8
  2. Very specific:

    Write the description in the language recent MRs of the repo use (glab mr list --merged -P 5 + glab mr view <n>).

    line 13
  3. Very specific:

    Run the title and description through the "AI-writing tells" checklist in <hub>/local/owner-prefs.md (Writing) and fix every hit before the …

    line 16
  4. Very specific:

    Repo context: read the repo's CLAUDE.md and its skills.

    line 20
  5. Very specific:

    Bug hunt: start the code-review skill on the branch diff against the target (it runs in the background).

    line 21
  6. Very specific:

    Triage: every code-review finding: valid → fix on the branch.

    line 23
  7. Very specific:

    Read the project's reviewers list in the hub registry (<hub>/local/projects.yaml.

    line 32
  8. Distinctive:

    Never merge a work MR. Never add AI attribution anywhere.

    line 38

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

MR

Every MR/PR you open for the user follows this flow. Never skip a step to save time.

1. Prepare (no forge writes yet)

  • Branch is pushed or ready to push, never main; commits follow the repo's message convention.
  • Find the repo's MR/PR template: .gitlab/merge_request_templates/Default.md (GitLab) or .github/pull_request_template.md (GitHub). Fill it in. Keep every heading; in checklists keep only the sections that apply, tick only what is actually true.
  • Write the description in the language recent MRs of the repo use (glab mr list --merged -P 5 + glab mr view <n>). The template and past MRs win over generic language rules.
  • Reference the ticket the way recent MRs do (e.g. Closes KEY-123).
  • Title: match recent merged MRs (same convention as commits, one ticket key, short).
  • Run the title and description through the "AI-writing tells" checklist in <hub>/local/owner-prefs.md (Writing) and fix every hit before the gate.

1b. Gate (before showing anything, and before every push of new commits)

Runs on its own: the user never has to ask for it. Same gate for a new MR, review fixes (review-threads) and any later push.

  1. Repo context: read the repo's CLAUDE.md and its skills. Untracked .claude/ dirs exist only in the main checkout, not in worktrees: read <repo path>/.claude/skills/ and <hub>/local/skills/<project>/ from there.
  2. Bug hunt: start the code-review skill on the branch diff against the target (it runs in the background).
  3. Patterns: meanwhile run the review skill in self mode: bin/crow-review-check for the regex rules, then the repo checklist + every other ledger row, for real.
  4. Triage every code-review finding: valid → fix on the branch; otherwise one line on why not (pre-existing, out of scope, wrong). If the code changed, re-check the new diff.
  5. A finding neither the ledgers nor the checklist would have caught gets a ledger row (self-review source).

Report: Patterns: N checked, K hits, M n/a, the code-review findings with their verdict, and what was fixed. Hits that need a decision become questions.

2. Validate with the user

Show the gate result, then the title and the description, and wait for an explicit OK. Adjust and show again if asked.

3. Ask for the reviewer

Read the project's reviewers list in the hub registry (<hub>/local/projects.yaml; hub path from $CROW_HOME, else ask). Ask which one with AskUserQuestion. Personal repos with no reviewers: skip.

4. Create

  • GitLab: glab mr create --source-branch <b> --target-branch main --title ... --description ... --assignee <user's forge username> --reviewer <chosen> --yes (add --draft if the user wants it or if pre-merge steps are pending).
  • GitHub: gh pr create --title ... --body ... --assignee @me [--reviewer <chosen>].
  • The user's forge username is the project's forge_user in the registry.
  • Never merge a work MR. Never add AI attribution anywhere.

5. Update the tickets

For every ticket the MR closes, move it to the tracker's review status (e.g. Jira "Merge Review"). A ticket the MR only partly covers stays in progress. Do the same when an existing MR changes state (ready, merged → the post-merge status the team uses).

6. Report

Give the link, the reviewer and any pre-merge steps left.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.