harness:verification-before-completion

Run precise proof commands before reporting completion with evidence.

3|1|Updated Apr 5, 2026
One-click install
npx skills add https://github.com/huangbaixun/harness-engineering --skill harness-verification-before-completion
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: harness:verification-before-completion
Source: https://github.com/huangbaixun/harness-engineering/tree/main/skills/verification-before-completion
Command: npx skills add https://github.com/huangbaixun/harness-engineering --skill harness-verification-before-completion

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Prevents premature or dishonest completion claims by requiring fresh verification evidence before you declare work complete, passing, fixed, or ready.

Core Features & Use Cases

  • Evidence-first completion gate: Forces you to identify the exact command that proves completion, run the full verification, and only then communicate success.
  • Gap-aware verification reporting: Blocks completion when any acceptance criterion is unsatisfied and reports the specific gap instead of general confidence.
  • Layer and architecture guardrails: Ensures verification includes required architecture-layer checks and flags reverse-direction dependency violations.
  • Use case: Before merging or archiving a change that claims rate-limit behavior, run the required verification and report whether each acceptance criterion (for example Retry-After header) is actually satisfied.

Quick Start

Ask your AI to run verification for the current building feature and only report “verified” with explicit evidence for every acceptance criterion.

Frequently Asked Questions about harness:verification-before-completion

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

FAQPage Schema
How do I verify evidence before claiming a pull request is ready?

Verify evidence before claiming done by identifying the exact proof command, running the full test suite, and only reporting success after fresh evidence confirms every acceptance criterion is satisfied.

What is evidence-first completion in agent workflows?

Evidence-first completion in agent workflows is an enforcement rule that blocks agents from asserting work is done by rejecting satisfaction language until mandated verification commands have actually been executed.

How do I enforce architecture layer dependency checks before finalizing work?

Enforce architecture layer dependency checks by running mandated integrations during finalization, which flag reverse-direction dependency violations and validate features.json acceptance criteria before allowing completion claims.

Why does my agent report work is fixed without running test evidence?

Agents report work is fixed without test evidence because they lack enforcement rules; this skill prevents dishonest completion claims by requiring fresh verification output before accepting any success language.

How do I validate features.json acceptance criteria before transitioning building to done?

Validate features.json acceptance criteria by running mandated integrations during the building-to-done transition, blocking completion and reporting the specific gap if any criterion remains unsatisfied.