agentleFS
OpenClawAndroidOpenClawAndroid/openclaw-android-assistant/.agents/skills/openclaw-pr-maintainer/SKILL.md

openclaw-pr-maintainer

Skill556 starsChanged 35 days agoMITView on GitHub

Review, triage, repair, or land OpenClaw issues and pull requests with current-source evidence and the native maintainer workflow.

15 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use openclaw-performance and plugin-sdk-impact.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • clawdbotSkill

    openclaw-performance

    Performance work in OpenClaw: a slow or heavy flow (latency, CPU, memory, SQL, RPC, renders), writing or reviewing a perf PR, or adding a cache. Count the work, delete waste first, cache last.

    Also in 7 projects that use this

  • clawdbotSkill

    plugin-sdk-impact

    Count Plugin SDK consumers across live ClawHub releases and bundled OpenClaw plugins before deprecation, removal, or behavior changes; supports public release-date filters.

    Also in 7 projects that use this

The file

Show its rules and text

We split this skill into the rules it gives an agent

17 rules. 0 are common, 13 are distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 13 edited versions we read keep these.

  • 13

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use for the requested OpenClaw issue/PR operation.

    line 8
  2. Distinctive:

    Review, triage, or list: read-only. Explain the change, actionable findings, useful evidence, and material uncertainty.

    line 14
  3. Distinctive:

    Keep checking pending requests and repair CI failures or conflicts under the same authority.

    line 21
  4. Distinctive:

    Queue/discovery: read triage. Batch live reads and shared discovery.

    line 25
  5. Distinctive:

    Author or regression attribution question: read author and provenance research.

    line 27
  6. Distinctive:

    Do not create user-owned app tasks to hide internal subtasks.

    line 33
  7. Very specific:

    Start with git status -sb.

    line 38
  8. Distinctive:

    Check assignment before deep work.

    line 51
  9. Distinctive:

    Read the affected owner, entry points, callers, relevant siblings, tests, docs, history, and current main as needed to support a conclusion.

    line 60
  10. Very specific:

    Use $openclaw-testing to choose proportional proof and its source-trust/host route.

    line 80
  11. Distinctive:

    Never describe mocks, skipped checks, or an older head as live evidence.

    line 85
  12. Very specific:

    Before committing or landing nontrivial code, run $autoreview and resolve verified actionable findings.

    line 97
  13. Distinctive:

    Use the current PR template.

    line 105
  14. Distinctive:

    Before public mutation, verify author/committer and authenticated writer identity.

    line 114
  15. Distinctive:

    Use body files for shell-sensitive text.

    line 115
  16. Distinctive:

    Verify the final merge/closure state and source rather than trusting a local summary.

    line 120
  17. Very specific:

    Never remove the worktree containing a live session's cwd, including your own.

    line 127

“Common” is measured: kept by at least 6 in 10 of the 13 edited versions of this file we could read. “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

OpenClaw PR Maintainer

Use for the requested OpenClaw issue/PR operation. Root AGENTS.md owns repair, product, source-trust, and approval policy; this skill owns the GitHub workflow. A pasted ref is context, not permission to publish or expand the task.

Select the scope

  • Review, triage, or list: read-only. Explain the change, actionable findings, useful evidence, and material uncertainty. Do not edit, assign, label, comment, or close unless separately authorized.
  • Fix only: investigate, repair locally, and validate. Publishing still requires ship/land or equivalent scoped authority.
  • Land/ship or autonomous repair: finish the authorized scope through current source proof, review, native merge submission, and verified remote merge. Keep checking pending requests and repair CI failures or conflicts under the same authority; accepted auto-merge is pending work, not task completion. Preserve contributor credit and unrelated work. Routine steps need no repeated permission; security, schema, product, release, and other root gates remain.
  • Queue/discovery: read triage. Batch live reads and shared discovery; do not turn one named PR into an unsolicited queue sweep.
  • Author or regression attribution question: read author and provenance research. Account age and contribution-history queries are conditional, not a prerequisite for every PR.

The lead remains hands-on. Use bounded independent workers where they add useful evidence; serialize shared checkout/ref mutations and GitHub writes. Do not create user-owned app tasks to hide internal subtasks. Preserve a stalled worker's patches, claims, and proof before taking over or reassigning its work.

Inspect the actual item

Start with git status -sb; preserve unrelated changes. Use available local gitcrawl data for related work when useful, then bare PATH gh with narrow JSON fields for live decisions. Stale/missing archives fall through to gh; do not broadly sync archives just to begin. PR source comes from gh pr view/diff and the checkout, not web search.

gh pr view <pr> --repo openclaw/openclaw \
  --json number,title,state,body,author,assignees,baseRefName,headRefName,headRefOid,files,comments,reviews
gh issue view <issue> --repo openclaw/openclaw \
  --json number,title,state,body,author,assignees,comments

Check assignment before deep work. Mention another assignee and treat a fresh assignment as active ownership unless the user directs otherwise. Assignment alone is not a block when stale or explicitly overridden. Claim an unassigned named item only under assignment or landing/autonomous-resolution authority, using the verified authenticated login. Never remove others' assignments without direction. Account/activity research belongs to its conditional route.

Establish the outcome

Read the affected owner, entry points, callers, relevant siblings, tests, docs, history, and current main as needed to support a conclusion. Inspect dependency contracts directly when they determine behavior; apply the root's personal Codex-source inspection requirement when relevant. Investigate contrary evidence before defending a verdict. Do not fill an arbitrary evidence matrix or invent a rejected alternative merely to satisfy a form.

  • Already fixed: prove the original behavior is repaired on current main, with the canonical commit/PR and matching source or runnable proof. Under closure authority, comment that evidence and close; otherwise report it.
  • Real defect: reproduce it, repair its owning invariant, and prove the repaired path and affected siblings. Prefer a coherent simplification over a symptom guard. Size and LOC are context, not correctness gates.
  • Existing useful PR: improve the editable branch rather than replacing it unnecessarily. If it cannot safely be updated, create an authorized credited replacement before closing the source. Never squash a contributor-owned PR after replacing its ancestry; use a maintainer-owned replacement.
  • Uncertain or product decision: state the concrete gap. Do not merge a speculative repair or automatically close work as out of scope.

Use $openclaw-testing to choose proportional proof and its source-trust/host route. Trusted development proof defaults local; remote proof needs an environment or isolation reason. Unavailable optional live proof may use meaningful boundary proof with the limitation stated. Explicit live requests, external API contracts, and changes whose risk requires authenticated execution keep their required proof. Never describe mocks, skipped checks, or an older head as live evidence.

CODEOWNERS review

CODEOWNERS routes review; check live GitHub enforcement. Restricted/security paths and material product, behavior, security, or ownership changes need listed-owner involvement. For ownership/review governance, verified active organization-admin direction also qualifies; repository admin/bypass alone does not. Neither route waives enforced reviews.

Review and publish

Before committing or landing nontrivial code, run $autoreview and resolve verified actionable findings. Use it for a review-only request when that request asks for independent autoreview; ordinary findings-only review does not need a second reviewer by default. Do not add another mandatory CLI review pass. Reopen review for substantive changes or unresolved concerns, not a patch-identical rebase or a mechanical head change. Address real human/bot findings and explain rejected ones; bot scores and Rank-up lists do not create separate obligations.

Use the current PR template. Lead with the plain-language problem and concrete user impact; keep the explanation short and leave implementation inventories in the diff or optional details. Keep important risks, migrations, required actions, useful evidence, known gaps, and contributor credit visible and current. Do not invent a user benefit for internal-only work. Explain material tradeoffs when they matter; do not require universal LOC tables, provenance fields, or alternate-fix essays. CHANGELOG.md is release-owned; user-facing release-note context stays in the PR/commit. Omit agent transcripts unless explicitly requested.

Before public mutation, verify author/committer and authenticated writer identity. Use body files for shell-sensitive text. Post findings only under the authorized write scope, and update an existing relevant comment/body instead of duplicating it. Publication, media, exact-head CI, and native landing mechanics are in landing; read that reference before publishing or merging.

Verify the final merge/closure state and source rather than trusting a local summary. Report the problem, owner-level change, important proof and limitations, credit, and linked final state in concise prose. Record worthwhile follow-ups; do not manufacture another task after a bounded request is complete.

Finalize and clean up

Never remove the worktree containing a live session's cwd, including your own. A child command's cd or cwd override does not move the parent session. Defer that worktree's removal until the session exits or releases its cwd; report the retained path in the final handoff. Once the requested outcome and required verification are complete, remove the task's disposable proof and scratch. This includes test logs, receipts, local proof archives, and .crabbox outputs. Do not retain, archive, export, or hand off these files merely to remove the completed task's worktree. Follow native closeout.

Required pre-merge proof stays mandatory. Preserve requested deliverables, explicit retention requests, unfinished source, recovery state needed by unfinished operations, active owners, credentials, agent state, and shared dependencies. Optional follow-ups do not keep a completed task open.

This applies to every task, not only PR work: report routine findings in chat/stdout. Create files only for deliverables or concrete tool/proof/recovery needs; state their purpose and reuse them. After the requested outcome and required verification are complete, remove task-owned proof, scratch, and finished worktrees through their native lifecycle. Do not require an archive or evidence handoff. Preserve requested deliverables, explicit retention requests, unfinished source and recovery state, live owners, credentials, agent state, shared dependencies, and unknown ownership.

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.