Actual-sync: Command for Claude Code

Gate a GitHub ticket for readiness

Run the ticket readiness gate on a GitHub issue before implementation begins.

Formerly /review-ticket — renamed to /gate-ticket to match the forge-kit convention used across the other projects. Update any muscle memory; the old name no longer resolves.

Usage

/gate-ticket <issue-number>

Argument: GitHub issue number (required). Example: /gate-ticket 44

What this does

Invokes the ticket-gate agent, which validates the issue through 3 sequential domain specialist agents. (This project runs a deliberately lean 3-agent panel rather than forge-kit's 5 core agents — see the agent file's "Scope" section for why, and for what manual review to add when a ticket touches security/privacy.)

1. actual-api — verifies correct @actual-app/api lifecycle, method names, field names, and known quirks
2. qa — verifies test cases are specific and cover error paths, shutdown assertions, and coverage thresholds
3. release-manager — verifies PR scope, commit convention, version bump, and measurable acceptance criteria

Each agent scores the issue 1–10. All three must score 10/10 for the issue to PASS.

Output

The gate posts a scorecard as a GitHub comment on the issue and reports:

• PASS — implementation may begin
• BLOCKED — exact list of required changes; re-run after addressing them

Steps

Use the Agent tool with subagent_type: ticket-gate, passing the following prompt:

Run the ticket readiness gate on GitHub issue #<issue-number> in agigante80/Actual-sync.

Follow all steps in the ticket-gate agent instructions:
1. Fetch the issue with gh issue view
2. Load CLAUDE.md, src/syncService.js (API call sequence), and src/__tests__/helpers/testHelpers.js
3. Thin-ticket pre-check (with CLAUDE.md context loaded — conventions the docs answer are
   not gaps); if 3+ material unanswered questions remain, post them as a comment and halt
   BLOCKED without scoring
4. Run complexity research (if triggered) and the codebase-exploration sub-agent (only when
   2+ complexity signals fire or the area is unfamiliar; skip for simple tickets)
5. Run actual-api agent — wait for result — then qa agent — wait for result — then release-manager agent.
   IMPORTANT: one agent per message, strictly sequential, never in parallel.
6. Compile the scorecard (include Research performed / Codebase context sections)
7. Post it as a GitHub comment
8. Report PASS, or on any score < 10 auto-remediate: write the updated body to a file and
   apply it with `gh issue edit --body-file` (never `--body "..."`), replacing rather than
   duplicating any prior "Required additions" section per agent; report BLOCKED