agentsclimarketplace

Loopify review

Skill AndreRatzenberger/loopify/plugins/loopify/skills/loopify-review

Loop yes. Prompt no. Agent skills that turn specs into checked agent loops with receipts.

Install
npx -y skills add AndreRatzenberger/loopify --skill loopify-review

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

One thing to look at

  • 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

Use when auditing a Loop Contract, trace, quality gate, or done claim — or when acting as the independent verifier that grades a loop's success claim and writes verdict.md.

SKILL.md

2.4 KB, as published. Nobody here has run it

loopify-review

Review a loop before or after execution.

When To Use

  • A loop claims done.
  • A quality gate may be incomplete.
  • A contract may have weak evidence or broad authority.
  • A trace needs review before a human accepts the result.
  • A loop requests verification per its contract's ## Verification section (verifier mode).

Do Not Use

  • Do not rubber-stamp passing tests.
  • Do not bury findings below a summary.
  • Do not treat missing evidence as low risk by default.

Workflow

  1. Read Loop Contract, trace, checks, and diff if present.
  2. Review requirement coverage using references/review-rubric.md.
  3. Identify weak evidence, missing checks, ambiguous stops, authority creep, and untested behavior.
  4. Present findings first, ordered by severity.
  5. Recommend concrete contract/check/trace fixes.

Verifier Mode

When invoked as the independent verifier (never the maker's session):

  1. Read references/independent-verification.md and the loop folder's contract, trace, and diff.
  2. Re-run the quality gate and the contract's cheap checks yourself.
  3. Grade every Requirement Evidence Map row pass/fail/cannot-verify.
  4. Confirm the manual/visual queue is enumerated.
  5. Produce verdict.md from templates/verdict.md. In a writable context that file is your only write; in a read-only sandbox (the emitted Codex verifier) return the full verdict body instead, and the caller transcribes it verbatim into the loop folder.
  6. Overall: approve only when every automated row passes and nothing blocking remains; otherwise reject with severity-ordered findings, or cannot-verify with what was missing.

Output

A review report using templates/review-report.md.

Validation

  • Findings are severity-ordered and path-grounded when possible.
  • Residual risk is explicit.
  • The verdict states whether the work is allowed to count.
  • Verifier mode: verdict.md exists, names the verifier and independence level, and every Overall=approve is backed by re-executed checks.

Example

Use loopify-review on .loopify/loops/001-papertrail/loop-contract.md, .loopify/loops/001-papertrail/trace.md, and the current git 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.