agentleFS
PorabuildPorabuild/Poracode/.agents/skills/finalize/SKILL.md

finalize

Skill121 starsChanged 5 days agoApache-2.0View on GitHub

Finalize the current Poracode worktree with deep-review, manual QA, issue fixes, verification, commit, push, PR creation or update with screenshots where they help review, and merge after checks pass. Use when the user says Finalize or requests the complete workflow. Creating this skill does not invoke it.

22 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use simplify, debug and 1 more.

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.
  • Hugging FaceSkill

    simplify

    Refactor code after tests pass. The "Refactor" phase of Red-Green-Refactor.

    Also in 3 projects that use this

  • AnthropicSkill

    debug

    Structured debugging session — reproduce, isolate, diagnose, and fix. Trigger with an error message or stack trace, "this works in staging but not prod", "something broke after the deploy", or when behavior diverges from expected and the cause isn't obvious.

    Also in 3 projects that use this

  • MicrosoftSkill

    handoff

    Compact the current conversation into a handoff document for another agent to pick up.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

7 rules. 0 are common, 2 are distinctive, and 5 are very specific to this project.

  • 0

    Common

    Most of the 21 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 5

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If the target or ownership is genuinely ambiguous, resolve it before publication.

    line 10
  2. Very specific:

    Read repository instructions and inspect the full intended diff against origin/master, including committed, staged, unstaged, and untracked …

    line 12
  3. Distinctive:

    Keep a compact record of completed checks, their commands/results, and the tree or affected files they covered.

    line 14
  4. Very specific:

    Run deep-review in fix mode.

    line 16
  5. Very specific:

    Use interactive-testing for real manual QA of affected app behavior, with an isolated managed session.

    line 17
  6. Very specific:

    Use commit, push the intended branch, and use pull-request to create or update its PR against master.

    line 34
  7. Very specific:

    Read fresh exact-number PR state.

    line 44

“Common” is measured: kept by at least 6 in 10 of the 21 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

Finalize

The user's invocation authorizes this entire workflow, including readiness and merge. Honor narrower instructions. Stay in the current worktree and keep its existing non-master branch. Preserve unrelated changes and staged work. If the target or ownership is genuinely ambiguous, resolve it before publication.

  1. Read repository instructions and inspect the full intended diff against origin/master, including committed, staged, unstaged, and untracked work. Record the starting branch, HEAD, and paths. Keep a compact record of completed checks, their commands/results, and the tree or affected files they covered. Reuse evidence from this session when its inputs are unchanged; committing alone does not invalidate it.
  2. Run deep-review in fix mode. Validate findings and fix in-scope issues without unrelated cleanup.
  3. Use interactive-testing for real manual QA of affected app behavior, with an isolated managed session. Add provider-chat-smoke or mobile-apps when applicable. Record real interactions and outcomes; tests do not replace manual QA. A diff that ships no runtime code or assets (docs, agent instructions, CI workflows, tests, or tooling that leaves the built app unchanged) has nothing to exercise: launch no app and record QA as not applicable, with the reason. Before launching, choose the screenshot level defined in pull-request and take its shots in this QA session once the UI is final. Only the before/after level needs a baseline worktree, install, and session; skip them otherwise. When that pass finishes, delete its smoke session directory, as interactive-testing requires. Finish heavy app builds before starting local tests; avoid overlapping baseline/branch builds and test runs. Reuse valid captures and managed builds while their relevant source, fixtures, and runtime inputs are unchanged.
  4. Fix confirmed QA issues, then run the full verify gate and every applicable conditional gate on the final tree. Because step 7 waits for CI, run the focused tests instead of the local bun run test, as verify allows; run the full suite locally only when the user asks. Run the gate once, then rerun only failed checks or checks invalidated by subsequent changes. A docs-only edit requires its applicable format/skill validation, not another app build, UI capture, or product test suite. Diagnose failures before retrying. For test failures, rerun the complete failing files after resolving the cause; for host contention, stop owned competing work and use supported worker limits without changing test thresholds. Preserve and report the original failures and retry results. Unavailable required QA or an unresolved red gate blocks merge.
  5. Use commit, push the intended branch, and use pull-request to create or update its PR against master. Finalize requests a ready PR. Include final verification/QA evidence and the screenshots the chosen level calls for. How verified lists the local checks and whether CI runs the full test suite for this diff (scripts/ci-change-scope.mjs decides).
  6. Monitor checks for the latest pushed SHA. Read failure logs, fix in-scope failures, verify, create follow-up commits, push, and refresh the PR description as needed. For a failed test shard, run its failing files locally with bun run vitest run <files> before and after the fix instead of the full suite. Every push requires a fresh CI assessment. Confirm suspected unrelated failures on unchanged origin/master; do not change unrelated code or bypass checks. Use CI status/watch operations between meaningful state changes; do not rerun valid local gates while waiting.
  7. Read fresh exact-number PR state. Verify the PR head equals the intended pushed commit, all applicable checks passed, required reviews are satisfied, and GitHub permits the merge. Empty/pending/cancelled checks are not green. Merge with the configured method, expectedHeadSha set to the verified SHA, and without admin override. If the internal tool lacks that field, use gh pr merge <number> --<method> --match-head-commit <verifiedSha>. Read back until GitHub confirms merged. Missing screenshots that the chosen level requires also block merge.
  8. Refresh Poracode's state with gh_get_pr and git_status for the same project/worktree. Prefer its internal git_* and gh_* tools throughout. Use CLI only when tools lack an operation, including PR attachment upload or title/body edits; follow each with an internal read to synchronize the UI. On older hosts, inspect tool schemas and use exact-number CLI readback if the result omits core PR state.

Continue autonomously through ordinary failures. Stop for a concrete external blocker, ambiguous scope, required human review, or a repair that fails twice without a new diagnosis. State the PR URL, SHA, failing requirement and next step. Never force-push, weaken tests, skip hooks, or delete the current worktree. Report the confirmed merge and concise verification/QA evidence; do not claim completion while GitHub only queued the merge.

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.