agentsclimarketplace

Finish task

Skill fbarrento/laravel-agent-kit/skills/finish-task

Laravel conventions, as agent skills. Your team's rulebook — portable across Claude Code, Codex, Cursor, and anything else SKILL.md.

Install
npx -y skills add fbarrento/laravel-agent-kit --skill finish-task

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Run a project's quality gates, get user review, then commit — the end-of-task counterpart to new-task. Tech-stack agnostic — discovers the project's own test, typecheck, lint, static-analysis, and format commands rather than hardcoding them. Use when a coding task is implemented and ready to commit, when the user says "finish this", "wrap up", "ready to commit", "ship it", or before any commit that concludes a task.

SKILL.md

3.0 KB, as published. Nobody here has run it

Finish Task

End-of-task lifecycle: gates → review → commit. Counterpart to new-task (which starts the branch). Commit/PR naming is owned by git-conventions.

This skill owns the gate sequence and policy, never the commands — those are stack-specific and get discovered from the project.

Gate sequence

Run in this order; stop on first red and enter the fix-loop:

  1. Tests
  2. Typecheck
  3. Lint
  4. Static analysis
  5. Format (apply, then re-stage)

Then: show the user the diff + gate results → ask for review → commit only on explicit approval.

Command discovery (stack-agnostic)

Do not assume tools. Detect the project's own commands, in priority order:

  1. Composer scripts — composer.json scripts (e.g. test, lint, analyse).
  2. NPM/PNPM scripts — package.json scripts.
  3. Makefile / justfile targets.
  4. bin/ or scripts/ entrypoints.
  5. CI config (.github/workflows/*, .gitlab-ci.yml) — mirror what CI runs.

Map each gate to the discovered command; skip a gate with no command (state which were skipped). Example resolutions (illustrative, not hardcoded): Laravel → pest/phpunit, phpstan/larastan, pint/rector. JS/TS → vitest/jest, tsc --noEmit, eslint, prettier.

Laravel-specific gate commands live in laravel-rules; this skill resolves "run the static-analysis command" → that project's tool.

Fix-loop

Gate red → read failure → fix → re-run that gate. Cap at ~3 iterations per gate; if still red, stop and escalate to the user with the exact error. Never disable/skip a check to make it pass. Never --no-verify.

Commit

  • Only after all runnable gates green AND user approved the diff.
  • Message per git-conventions (type vocab, subject grammar, footers).
  • Push only if the user asks.

Guardrails

  • Never commit red — no failing or skipped-to-pass checks.
  • Never commit without showing the diff and getting approval — the commit concludes the task; treat it as a confirmation gate.
  • Never bypass hooks (--no-verify), never weaken a check to pass.
  • Never commit secrets — scan the diff for tokens/keys before committing.
  • If the working tree mixes unrelated changes, ask before staging.

Checklist

  • Gate commands discovered from the project, not assumed.
  • Each gate run or explicitly reported as skipped (no command).
  • All runnable gates green before commit.
  • Diff shown + user approved.
  • Commit message follows git-conventions.
  • No hook bypass, no weakened checks, no secrets in the diff.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.