gate-check

Dispatches a sub-agent to run quality gates and verify acceptance-criteria test coverage for a ticket worktree.

1|Updated Jul 15, 2026
One-click install
npx skills add https://github.com/stefanrusek/dark-harness --skill gate-check-stefanrusek
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: gate-check
Source: https://github.com/stefanrusek/dark-harness/tree/main/.claude/skills/gate-check
Command: npx skills add https://github.com/stefanrusek/dark-harness --skill gate-check-stefanrusek

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Verifying that a finished ticket actually passes the project's quality gates is easy to shortcut when the coordinator runs checks inline and informally. This Skill turns gate verification into a strict, checked-in, auditable procedure executed by a dedicated sub-agent, so every verdict follows the same rules defined in CLAUDE.md. ## Core Features & Use Cases - Real gate execution: Runs the exact CLAUDE.md §5 commands (typecheck, lint, test:coverage, e2e) from the ticket's worktree and records pass/fail per gate. - Acceptance-criteria verification: Maps every User Story bullet to a named test file and test case per CLAUDE.md §9, failing any criterion without located test evidence. - Ticket lifecycle integration: Transitions the ticket to verifying via spile-ops, appends a dated Notes entry with the verdict, and commits only the ticket update. - Use Case: An implementer reports ticket DH-0113 as done. You fill the prompt template with the worktree path and ticket ID, dispatch it via forked-subagent, and receive a strict PASS/FAIL report with per-gate results and coverage evidence. ## Quick Start Fill the GATE-CHECK-PROMPT.md template with the worktree path and ticket ID, then dispatch it to a sub-agent using the forked-subagent script pointed at the ticket's existing worktree.

Frequently Asked Questions about gate-check

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

FAQPage Schema
How do I run quality gates on a finished ticket worktree?▼

Fill the GATE-CHECK-PROMPT.md template with the worktree path and ticket ID, then dispatch it to a sub-agent, preferably via the forked-subagent script. The sub-agent runs typecheck, lint, test:coverage, and e2e exactly as CLAUDE.md §5 specifies and reports a strict verdict.

How do I verify acceptance criteria have test coverage?▼

The gate-check sub-agent reads the ticket's User Stories section and locates a specific test file and test case for every Given/When/Then bullet. Any criterion without a named, located test is reported as uncovered and fails the overall verdict.

Does the gate-check sub-agent fix failing tests or gates?▼

No. The sub-agent only reports failures and never patches code, adds tests, or modifies source files. A FAIL verdict is routed back to the ticket's implementer, and its only write side effect is the ticket's Notes entry and status transition.

When should I dispatch gate-check instead of running gates inline?▼

Dispatch it whenever an implementer reports a ticket's work as done and it needs verification before moving from implementing to verifying. Running gates inline as the coordinator is informal, unrecorded, and easy to shortcut under time pressure.

What happens if the worktree path is wrong or the ticket file is missing?▼

The sub-agent stops and reports a hard failure rather than guessing. It does not fall back to another directory if the worktree is invalid, and it halts if zero or multiple ticket files match the given ticket ID.