verification-before-completion

Enforces running verification commands before claiming work is complete or passing.

Updated Jul 29, 2026
One-click install
npx skills add https://github.com/MaiconGambini/opencode-harness-guide --skill verification-before-completion-maicongambini
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: verification-before-completion
Source: https://github.com/MaiconGambini/opencode-harness-guide/tree/main/skills/verification-before-completion
Command: npx skills add https://github.com/MaiconGambini/opencode-harness-guide --skill verification-before-completion-maicongambini

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? AI agents and developers often claim work is complete, fixed, or passing without actually running the verification commands, leading to broken builds, shipped bugs, and lost trust. This Skill enforces an evidence-first discipline: no completion claims without fresh verification output. ## Core Features & Use Cases - Verification Gate: A five-step gate function (identify, run, read, verify, claim) that must be completed before any success statement. - Failure Pattern Catalog: Tables mapping common claims (tests pass, build succeeds, bug fixed) to the exact evidence required and what is not sufficient. - Rationalization Blocking: Explicit counterarguments for excuses like "should work now", "linter passed", or "agent said success". - Use Case: Before committing code or creating a PR, the agent runs the full test suite, reads the output confirming zero failures, and only then states that tests pass with the evidence attached. ## Quick Start Before claiming any task is done, run the relevant verification command and report the actual output alongside your claim.

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 committing code?

Identify the command that proves your claim, run it fresh and in full, read the complete output including exit codes, and only then state the result with that evidence. Never claim success based on previous runs or assumptions.

What counts as sufficient evidence that tests pass?

Sufficient evidence is the actual test command output showing zero failures from a fresh run in the current session. Previous runs, partial checks, or statements like "should pass" do not qualify as verification.

Can I trust an agent's success report without checking?

No. Agent success reports must be independently verified by inspecting the VCS diff and confirming the changes actually exist and work. Trusting reports without verification is listed as a common failure pattern.

Why is linter output not enough to claim a build succeeds?

Linters check style and syntax but do not verify compilation. Claiming a build succeeds requires running the actual build command and confirming exit code 0, since linter passing proves nothing about compilation.

How do I verify a regression test actually works?

Use the red-green cycle: write the test, run it to confirm it passes, revert the fix and confirm the test fails, then restore the fix and confirm it passes again. A test that never fails proves nothing.