Nurture

Mode A — Triage (no args)

List every file in {CONTENT_DIR}/specs/drafts/. For each one:

1. Read the file
2. Print a one-line summary: [filename] — {title} ({source tag if present})
3. Ask the user: pursue, defer, or discard?

After each decision:

• pursue — add a status: pursue line to the stub's frontmatter (or top of file if no frontmatter) and move on
• defer — add status: defer and move on
• discard — confirm once, then delete the file

After all stubs are triaged, print a summary:

Triage complete — {N} pursue, {N} defer, {N} discarded
Pursue list: {names}
Run `/nurture {name}` to flesh out any of these.

Mode B — Flesh out a stub (arg = spec name fragment)

Find the matching file in {CONTENT_DIR}/specs/drafts/ by name fragment. Read it fully.

Then run a focused grill-me interview to build it into a complete spec. Follow the grill-me rules:

• One question at a time
• Walk every branch of the decision tree
• Provide a recommended answer for each question
• If a question can be answered by exploring the codebase, do that instead of asking

Branches to cover (skip any already answered in the stub):

1. Problem — what pain does this solve, for whom, how often?
2. Sources & inputs — what data/systems does it touch?
3. Output — what does done look like? Files, UI, API, side effects?
4. Stack — language, libraries, auth strategy
5. Phases — what's Phase 1 (smallest useful thing)?
6. Risk — what's the hardest part?
7. Prerequisites — any one-time setup needed?

When the interview is complete, rewrite the stub file as a full spec using the gestor-collector spec at {CONTENT_DIR}/specs/gestor-collector.md as a format reference.

Ask the user: "Move to candidates for /review-spec?"

• Yes → move file from {CONTENT_DIR}/specs/drafts/ to {CONTENT_DIR}/specs/candidates/
• No → leave in drafts