ghpm-issue

ghpm-issue

PREREQUISITE: Read ../ghpm-shared/SKILL.md for prerequisites and error handling.

File a GitHub issue and add it to the project board in one step. Works mid-session (during ghpm-work) or standalone.

Arguments

• /ghpm-issue — interactive: asks for title and body
• /ghpm-issue <title> — uses argument as issue title, asks for optional body

Workflow

Phase 1: Setup

1. Startup sequence: Follow the startup sequence in ../ghpm-shared/SKILL.md. This loads config, refreshes cache if stale, and checks for stale sessions.

   Extract from config:

   • config.project — id, owner, number, title
   • config.repos — eligible repositories
   • config.workflow — columns, field_id
   • config.fields — project fields with name, id, type, options
   • config.conventions — natural language rules (if present)

Phase 2: Gather Info

1. Title and body:

   • If argument provided, use as title. Ask for an optional body (user can skip).
   • If no argument, ask for title, then ask for optional body.

2. Pick repo:

   • If config.repos has one entry → use it.
   • If config.repos is empty or missing → ask the user to specify a repo in owner/repo format.
   • If config.repos has multiple entries → present a numbered list and ask.

3. Link to active session: Find active session per ../ghpm-shared/references/session.md (match current branch against session files). If found, prepend to the issue body:

   Related to <repo>#<num>
   
   <original body>
   

Phase 3: Infer Metadata

All inference is best-effort. Skip silently when no confident match exists.

1. Infer labels: Fetch the target repo's actual labels:

   gh label list -R <repo> --json name,description --jq '.[] | "\(.name)\t\(.description)"'
   

   Match the issue title and body against the fetched label names and descriptions (case-insensitive substring/keyword matching). Only use labels that exist on the repo — never assume a label exists.