agentsclimarketplace

Verify before done

Skill yuchi-chang/no-cape/skills/verify-before-done

Use when about to claim work is complete, fixed, or passing — before committing, opening a PR, marking a task done, or relaying a subagent's success reportFrom its SKILL.md

Install
npx -y skills add yuchi-chang/no-cape --skill verify-before-done

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.
  • 2 stars2 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.

SKILL.md

2.7 KB, 561 tokens by cl100k_base, as published. Nobody here has run it

Verify Before Claiming Done

Evidence before claims. If the verification didn't run in this turn, the claim can't be made.

The gate

  1. Name the command that would prove the claim.
  2. Run it fresh and completely — not a remembered earlier run, not a partial subset.
  3. Read the full output and the exit code.
  4. Only then state the result — with the evidence, or with the actual failing state.

What each claim requires

ClaimProof required
Tests passFresh full test run with 0 failures — lint or build passing is not it
Build succeedsBuild command exits 0 — lint passing is not it
Bug fixedThe original symptom re-tested and gone
Regression test worksRed-green verified once: revert fix → test fails; restore fix → test passes
Subagent finishedInspect the actual diff — never trust the agent's own "success" report
Requirements metRe-read the spec/plan and check items line by line — tests passing ≠ spec met

Check the criterion, not just the run

A green check only proves what it measured. Before trusting it, ask: does passing this actually mean the work is right? A test that asserts "ran without error" passes on output that silently dropped half its input. The gate above stops you skipping verification; this stops you trusting a verification that checks the wrong thing.

The trap is loudest when a library or default does the work for you — vad_filter, pagination limits, truncation, sampling, auto-retry-then-give-up. The call "succeeds" while quietly discarding part of the result. So when the output is a transformation of an input (transcribe, scrape, convert, migrate, extract), assert on coverage, not just liveness:

  • Not "got N segments" but "the N segments cover the input" — duration, word count, row count, page count in vs out.
  • Diff the risky default against its alternative once (e.g. filter on vs off) to see what it actually costs you — don't assume, measure.
  • If you can't state what complete output would look like, you can't yet claim the run proved anything.

Red flags

"Should work now", "probably passes", expressing satisfaction before running anything, committing without a fresh run. Asserting the pipeline ran without asserting the output is complete. Blaming a component (a filter, a default) for dropping data before reproducing it — name the cause from evidence, not a hunch. If verification fails, report the real state with the output — never optimism.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Gives 0 of the 12 instructions most quality gates skills give in 561 tokens

Counted across 1,195 of the 2,094 authors here whose files we hold, read 2026-08-07

  • Read the output and check the exit codein 54 of 1195, across 14 files
  • Verify requirements using a line-by-line checklistin 53 of 1195, across 12 files
  • Identify the verification command proving the claimin 51 of 1195, across 12 files
  • Run the full verification commandin 50 of 1195, across 11 files
  • Verify output confirms the claimin 49 of 1195, across 12 files
  • Check version control diff after agent delegationin 46 of 1195, across 6 files
  • State claim with evidencein 44 of 1195, across 4 files
  • Run the test suitein 33 of 1195, across 26 files
  • Keep state in memory by defaultin 27 of 1195, across 6 files
  • Make prototype runnable with one commandin 26 of 1195, across 5 files
  • Produce a verification reportin 25 of 1195, across 14 files
  • Detect the package manager from lockfilesin 24 of 1195, across 5 files

Said here and by no other author read

  • state results using fresh evidence
  • re-read the spec line by line to confirm requirements
  • verify the chosen criterion matches the claim
  • assert output coverage not just successful execution
  • measure the cost of risky defaults by diffing alternatives

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 326,871. 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.