GitHub issues
Record the user's problem and desired outcome in the requested repository. A request to create or update an issue authorizes that action; a request to draft or plan does not authorize publication. Inspect the target issue before updating it and preserve unrelated content and metadata.
Issue content
An ordinary issue usually needs only:
## Problem
Describe the current experience, who it affects, and why it matters.
## Desired outcome
Describe how that experience should improve.
Add observed versus expected behavior and reproduction evidence for a bug when available. A motivating example may illustrate a broader problem; do not silently narrow the scope to that example or speculate about an unverified fix.
Do not add architecture, file lists, scope/non-goals, implementation steps, acceptance checklists, tests, commands, or delivery phases by default. Include technical detail when the user asks for a specification or implementation task. Keep parent epics focused on the overall problem.
Publication and metadata
- Prefer a structured tool argument or
gh issue create --title "..." --body-file <path>for multiline text. Verify the destination repository from current context or GitHub configuration. - Inspect existing labels when assigning them. Reuse the current app/feature grouping and colors; do not infer a new label from an obsolete app name or create a taxonomy as a side effect of an ordinary issue request.
- Use GitHub's native fields for issue type, parent/sub-issues, and blocked-by dependencies. Avoid duplicating metadata in the body.
- Verify issue numbers before cross-referencing them. Local spec numbers and filenames are not GitHub issue numbers; keep a mapping when publishing a breakdown.
- Make the issue self-contained. Temporary local paths are not usable context for a GitHub reader.
- Confirm the created or updated issue and return its link. If a batch partly fails, report completed work and inspect live state before retrying.
For an explicit request to split work into multiple issues, use zoonk-issue-planning. For issue types, dependencies, or sub-issue API operations, load relationship recipes. A single ordinary issue does not require either workflow.