harness-verify

Runs the project's verify_all script and reports PASS, WARN, or FAIL results.

Updated May 15, 2026
One-click install
npx skills add https://github.com/Alan-IFT/harness-kit --skill harness-verify-alan-ift
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: harness-verify
Source: https://github.com/Alan-IFT/harness-kit/tree/main/skills/harness-verify
Command: npx skills add https://github.com/Alan-IFT/harness-kit --skill harness-verify-alan-ift

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When an AI agent claims a task is done, there is no objective proof the work actually passes the project's checks. This Skill runs the project's total verification script and returns a structured PASS / WARN / FAIL verdict so completion claims are grounded in real test and check results. ## Core Features & Use Cases - Single-command verification: Invokes .harness/scripts/verify_all.sh and captures stdout, stderr, and the exit code. - Structured verdicts: Maps exit codes to PASS (0), WARN (1), or FAIL (2) and prints a per-check summary with failure details. - Task history recording: Appends results to the active task folder's ## verify_all result section and to .harness/scripts/verification_history.log. - Use Case: After a developer agent finishes a feature, run verification before delivery; on FAIL, route the task back with the exact failing checks instead of declaring completion. ## Quick Start Ask the AI to run harness-verify to confirm the current task passes all project checks before marking it done.

Frequently Asked Questions about harness-verify

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

FAQPage Schema
How do I verify a task is done in a Harness project?▼

Run the harness-verify command, which executes .harness/scripts/verify_all.sh and returns PASS, WARN, or FAIL based on the exit code. A FAIL result means the task is not done and must not be declared delivered.

What do the PASS, WARN, and FAIL results mean in verify_all?▼

The exit code of verify_all.sh determines the verdict: 0 means PASS, 1 means WARN, and 2 means FAIL. On WARN the task may proceed with warnings logged; on FAIL the task is routed back for fixes.

What happens if verify_all.sh does not exist?▼

The Skill reports that the project is not Harness-initialized and suggests running /harness-init or /harness-adopt first. Verification cannot run without the script in place.

Can I skip slow checks like e2e tests during verification?▼

Yes, pass the --quick option to skip slow checks such as end-to-end tests. The flag is forwarded directly to the underlying verify_all script.

Why does verify_all FAIL even though the code looks correct?▼

FAIL indicates at least one check returned a failing exit code, such as unit test failures or missing agent definitions. The summary lists each failing check with details pointing to logs like scripts/last_run.log.