planwerk-agent

Appearance

Generate a fix/implement prompt ​

Generate a copy-paste-ready Claude Code prompt for an existing GitHub issue — either to fix an audit finding or to implement a proposal/elaborated issue. No Claude call is involved; the prompt is a deterministic assembly so the output is stable and safe to pipe into other tools.

bash
# Auto-detected mode (audit titles get the fix prompt, others the implement prompt)
planwerk-agent prompt https://github.com/owner/repo/issues/42

# Force the fix variant
planwerk-agent prompt --mode fix owner/repo#42

# Force the implement variant
planwerk-agent prompt --mode implement owner/repo#42

# Pipe straight into the clipboard (macOS)
planwerk-agent prompt owner/repo#42 | pbcopy

Mode auto-detection looks at the issue body: audit-generated issues carry a **Severity**: marker and get the "fix" prompt, everything else gets the "implement" prompt. See the CLI reference for the --mode flag.

How it works ​

  1. Issue Input: A GitHub issue reference (URL or owner/repo#number).
  2. Fetch Issue: Title, body, URL, and state are fetched via gh issue view.
  3. Mode Selection: auto (default) inspects the issue body — audit findings carry a **Severity**: marker and get the "fix" prompt; everything else gets the "implement" prompt. Override with --mode fix or --mode implement.
  4. Prompt Assembly: The runner deterministically assembles a prompt containing the agent workflow, rules (no scope creep, no --no-verify, run tests, update docs), and the issue metadata + body. No Claude call is made — the output is reproducible so it can be piped into other tools or diffed over time.
  5. Output: The prompt is written to stdout, ready to paste into Claude Code or any other AI coding agent.