verification-before-completion

Requires running verification commands and checking output before claiming work is complete.

1|Updated Jul 17, 2026
One-click install
npx skills add https://github.com/dariovr1/sixth-sense --skill verification-before-completion-dariovr1
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: verification-before-completion
Source: https://github.com/dariovr1/sixth-sense/tree/main/skills/verification-before-completion
Command: npx skills add https://github.com/dariovr1/sixth-sense --skill verification-before-completion-dariovr1

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? AI agents and developers often claim tasks are done, tests pass, or bugs are fixed without actually running the verification command, leading to false success reports and broken commits. This Skill enforces an evidence-first gate so no completion claim is made without fresh command output. ## Core Features & Use Cases - Verification Gate: A five-step process (identify, run, read, verify, claim) that must be completed before any success statement, commit, push, or PR. - Claim-to-Evidence Mapping: A table defining what output each common claim requires, such as test runs showing zero failures or builds exiting with code 0. - Corner-Cut Ledger: Greps the diff for cut-corner: markers and flags any without a revisit: trigger in the completion report. - Use Case: Before committing a bug fix, run the failing test, confirm it passes, check the exit code, and only then report the fix as verified with the actual output attached. ## Quick Start Before claiming this task is done, run the full test suite and show me the output proving zero failures.

Frequently Asked Questions about verification-before-completion

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I verify tests pass before committing code?

Run the full test command fresh in the current session, read the complete output, and confirm zero failures and a successful exit code. Only then claim the tests pass, citing the actual output as evidence rather than relying on a previous run.

What counts as evidence that a bug is fixed?

A bug fix is verified by re-running a test that reproduces the original symptom and confirming it now passes. Changing the code or assuming the fix works is not sufficient evidence without that fresh test output.

Can I claim work is done if I cannot run the verification command?

No. Name the specific check you cannot run and why, ask the user to provide access or run it themselves, and state the claim as unverified. Silently skipping an unreachable check is treated the same as skipping a reachable one.

Why do AI coding agents report false success on tasks?

Agents often extrapolate from partial checks, trust prior runs, or use hedging language like "should pass" instead of executing commands. This Skill blocks that pattern by requiring fresh command output before any success claim.

What is the corner-cut ledger check before completion?

It greps changed files for `cut-corner:` markers and flags any marker missing a `revisit:` trigger in the completion report. It is report-only, surfacing trade-offs so they stay visible rather than silently becoming permanent.