agentsclimarketplace

Verification before completion

Skill lawzava/megapowers/plugins/megapowers/skills/verification-before-completion

Skills, plugins, and hooks that change how coding agents work — Claude Code, Codex, OpenCode, Antigravity. Measured, not vibed.

Install
npx -y skills add lawzava/megapowers --skill verification-before-completion

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

  • 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.

What its author says it does

Copied from the file, not written here

Use before claiming work is complete, fixed, passing, ready to merge or publish, and before commits or pull requests. Triggers on any success or status claim.

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

3.4 KB, as published. Nobody here has run it

Verification Before Completion

Core principle: Evidence before claims, always.

This applies to the spirit of a completion claim, not just the exact wording. Rephrasing a success claim doesn't exempt it.

The Core Rule

Make no completion claim without fresh verification evidence.

If you haven't run the verification command in this message, you can't yet claim it passes.

The Gate

Before claiming any status or expressing satisfaction, work through these steps in order:

  1. Identify: What command proves this claim?
  2. Run: Execute the full command, fresh and complete.
  3. Read: The full output — check the exit code, count the failures.
  4. Verify: Does the output confirm the claim?
    • If no: state the actual status, with evidence.
    • If yes: state the claim, with evidence.
  5. Then make the claim.

Each claim has its own oracle. Tests need the test run, a build needs the build command, a bug fix needs the original symptom exercised, a regression test must fail with the fix reverted, and an agent's success report needs independent verification of the actual changes. Confidence, a partial check, or a passing linter is not that oracle.

Earned State

Every claim lands in one of three states. Claim only the highest one you earned:

  • Implemented: the requested change exists.
  • Locally verified: the canonical local oracle passes.
  • Externally verified: the real target environment and every required normal-user, external-service, or database witness pass.

A local test never silently earns external verification. For user-facing behavior, automated tests do not replace a normal-user witness: exercise the supported entry point with ordinary permissions and record the interaction, the visible result, and its provenance.

Acceptance Evidence Map

Work with several acceptance criteria, or any criterion needing an external, UX, or database oracle, earns a written map instead of prose tracking:

criterion | implementation target | local oracle | external, UX, or database oracle | earned state | evidence

Copy each criterion verbatim; never weaken or paraphrase one. For external database-backed behavior, record each cutpoint separately: caller request, service receipt and decision, target-environment database write or read, outward response, and user-visible result, each carrying its environment and correlation or record key. A missing required cutpoint leaves external verification pending or blocked.

A single-criterion local change does not need the table. It still needs the oracle, the output, and the honest state.

Unavailable Requirements

If a required dependency, tool, or input is unavailable, say so explicitly — a completion claim that hides a missing requirement is a false claim. A workaround or fallback never silently satisfies the requirement: disclose the substitution and report the requirement itself as unmet (blocked or partial) until the human accepts the substitute.

The gate holds before committing, opening a PR, marking a task done, moving to the next task, or delegating. Hedged wording ("should work now") signals a claim you have not yet earned the evidence for.

Origin: Derived from Superpowers (MIT, (c) 2025 Jesse Vincent), https://github.com/obra/superpowers.

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.