Skill: Refine Ticket
Skill: Refine Ticket
Automated issue body patching from review findings. Reads the review note, extracts corrections, applies them to the Forgejo issue body, and triggers a re-review. Eliminates manual translation between review findings and issue fixes.
Inputs
You receive from the router:
- Board item ID and board slug
- Review note slug — e.g.,
review-415-2026-03-27
Steps
- Read the review note —
mcp__pal-e-docs__get_note(slug="review-{id}-{date}"). Extract all findings from the Recommendation section. - Classify each finding:
- GET the current issue body — curl the Forgejo API to read the current body.
- Apply body fixes — for each body fix finding:
- PATCH the updated body — curl PATCH to Forgejo API with the corrected body.
- Apply label fixes — update board item labels via MCP.
- Post summary comment — on the Forgejo issue:
## Refinement Applied Review note: `review-{id}-{date}` Changes made: - {list of body changes} - {list of label changes} Flagged for Lucas: {any scope decisions needed} Re-review triggered. - Report results — return list of applied fixes, flagged items, and whether a re-review should run.
Review Note Format Requirement
For this skill to work automatically, review notes (from
skill-review-ticket) must include a structured ### Recommendation section. Each recommendation should be one of:[BODY]— fix in the issue body (e.g., "Fix file path: X → Y")[LABEL]— fix on the board item (e.g., "Add arch:hooks label")[SCOPE]— needs human decision (e.g., "Clarify: boot ordering or mid-uptime restart?")[DECOMPOSE]— needs sub-board (e.g., "7 AC across 3 systems, split into 3 tickets")
The refine agent reads these tags to determine what it can auto-fix vs what needs escalation.
The Automated Loop
Zero manual translation. Betty Sue invokes
/refine-ticket, the agent reads the review note, applies what it can, flags what it can't.Related
skill-review-ticket— produces the review notes this skill consumesskill-validate-ticket— the right-side equivalent (post-merge)template-board— used when [DECOMPOSE] recommendations are foundsop-board-workflow— column flow this skill accelerates