ghpm-issue（ghpm 问题）

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.

1. 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.

1. 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. If no labels match confidently, skip. Don't force a match.
2. Infer initial status: Use config.workflow.columns to determine the initial status:

- If config.conventions.status_sync is present, interpret it for new issue context (e.g., if it describes transitions from Planned, that implies new issues start at Planned or earlier). - Otherwise, use the first column in config.workflow.columns as the default. - Never hardcode a column name — always resolve from config.

1. Infer project fields: Iterate over config.fields entries where type is "single_select" and options is non-empty. For each field, infer the best option from the issue context (title, body, labels, repo name):

- Compare each options[].name (case-insensitive) against the issue context for a substring or keyword match. - If exactly one option matches confidently, use it. If ambiguous or no match, skip silently. - When in doubt, skip rather than guess.

Phase 4: Confirm and Create

1. Confirm: Prompt (create_issue): Show the resolved issue details — title, repo, labels (if any), initial status, inferred fields — and ask: "Create this issue and add to <project.title>?" Resolve via prompts.ghpm-issue.create_issue per ../ghpm-shared/references/prompts.md.
2. Create issue: gh issue create -R <repo> --title "<title>" --body "<body>" [--label <labels>] Extract the issue URL from stdout. Then fetch the node ID: gh issue view <url> --json number,url,id --jq '{number:.number, url:.url, nodeId:.id}'
3. Add to project: gh api graphql -f query=' mutation {addProjectV2ItemById(input: {projectId: "<config.project.id>" contentId: "<issue nodeId>"}) {item {id}}}' Extract the project item ID for field mutations. If this fails: The issue still exists on GitHub. Report the URL and provide the manual fallback: Could not add to project. Add manually: gh project item-add <config.project.number> --owner <config.project.owner> --url <issue-url>
4. Set status: Using the status resolved in step 6: gh api graphql -f query=' mutation {updateProjectV2ItemFieldValue(input: {projectId: "<config.project.id>" itemId: "<project item ID>" fieldId: "<config.workflow.field_id>" value: {singleSelectOptionId: "<resolved status column id>"}}) {projectV2Item {id}}}' If this fails, continue — the item is on the board, just without a status.
5. Set project fields: For each field inferred in step 7, set via GraphQL (run mutations in parallel where possible): gh api graphql -f query=' mutation {updateProjectV2ItemFieldValue(input: {projectId: "<config.project.id>" itemId: "<project item ID>" fieldId: "<field.id>" value: {singleSelectOptionId: "<matched option.id>"}}) {projectV2Item {id}}}' If a mutation fails, continue with remaining fields.

Phase 5: Output

1. Format per ../ghpm-shared/references/format.md: Created #<num> "<title>" in <repo> Project: <config.project.title> (<resolved status>) <field.name>: <option.name> (for each field set in step 12) URL: <issue-url>

Prompt Configuration

Per ../ghpm-shared/references/prompts.md. Config lives at prompts.ghpm-issue in .ghpm/config.json.

Action keys: create_issue.

Tips

• This skill does NOT create a work session. Use /ghpm-work <number> after filing to start working on the new issue.

See Also

• ghpm-shared — Prerequisites and error handling
• ghpm-work — Start a work session on an issue
• ghpm-suggest — Find what to work on next