verification-before-completion

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

Updated Aug 22, 2026
One-click install
npx skills add https://github.com/xegheplimo-web/hermes-ops --skill verification-before-completion-xegheplimo-web
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: verification-before-completion
Source: https://github.com/xegheplimo-web/hermes-ops/tree/main/skills/software-development/execution-discipline/verification-before-completion
Command: npx skills add https://github.com/xegheplimo-web/hermes-ops --skill verification-before-completion-xegheplimo-web

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Agents and developers often claim work is done, tests pass, or bugs are fixed based on assumptions, partial checks, or trusting another agent's report. This Skill eliminates false completion claims by requiring fresh verification evidence before any success statement. ## Core Features & Use Cases - Evidence Gate: A five-step gate function (identify, run, read, verify, claim) that blocks any completion claim not backed by freshly executed command output. - Claim-to-Evidence Mapping: A table of common claims (tests pass, linter clean, build succeeds, bug fixed) paired with the exact verification each requires and what is not sufficient. - Pipeline Integration: Hooks into the Hermes pipeline so Devin task completion, policy-gate merges, and Codex reviews all re-verify independently rather than trusting agent success reports. - Use Case: Before creating a PR, an agent runs the full test suite, confirms 0 failures in the output, and only then reports completion with the evidence attached. ## Quick Start Before claiming this task is done, run the full test and lint commands and show me the actual output as evidence.

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 work is complete before creating a PR?

Identify the command that proves the claim, run it fresh and in full, read the complete output including exit code and failure counts, then make the claim only with that evidence attached. Skipping any step means the claim is unverified.

How to validate that an AI coding agent actually finished its task?

Do not trust the agent's success report. Check the VCS diff to confirm changes exist, run the relevant test or build commands yourself, and report the actual verified state rather than the agent's claim.

What counts as sufficient evidence that tests pass?

Only fresh output from the test command showing zero failures counts. Previous runs, partial checks, statements like "should pass", or confidence without execution are explicitly not sufficient.

Does this verification gate work with automated CI pipelines?

Yes. It integrates with the Hermes pipeline where Devin includes verification evidence in PRs, the policy-gate re-verifies before merge, and Codex provides independent read-only review as an additional check.

Why do agents claim success when the code is broken?

Common rationalizations include "should work now", "I'm confident", or trusting another agent's report. The skill counters these by treating any claim without freshly run verification output as unverified, regardless of confidence.