shipping-and-archiving

{{ACTIVE_RULES}}

Shipping And Archiving

Overview

Formal completion of a design session. Verifies visual correctness one last time (Gate B), moves the CSpec to specs/shipped/, appends a history entry, and — when the design qualifies — extracts a reusable recipe back into the knowledge base.

When to Use

Invoke when the user:

• says "done", "ship it", "ship", "finish", "complete", "archive"
• has an active CSpec with a snapshot

Do NOT use if:

• the user is still iterating — use generating-figma-design or learning-from-corrections
• the user is abandoning the work — route to drop (handled by using-bridge command map)

Procedure

1. Auto-detect pending corrections

Check if the snapshot has been updated since the last fix run, or if fix was never run:

• If snapshot exists (specs/active/{name}-snapshot.json) AND no learnings reference this spec in learnings.json:

  • Re-extract current Figma state (same extraction as the learning-from-corrections skill, step 2)
  • Compare against snapshot
  • If changes detected:

    Changes detected since last generation/fix.
    Running auto-fix to capture corrections before archiving...
    

    -> Execute the full fix flow via the learning-from-corrections skill (steps 2-10 of its procedure)
  • If no changes: proceed

• If no snapshot exists: skip (design was never generated, spec-only work)

2. Final check

• CSpec exists in specs/active/
• User has validated the design (explicit confirmation)
• If Figma design exists: matches CSpec intent

3. Recipe extraction check

Evaluate if this design qualifies for recipe extraction:

Eligibility criteria (ALL must be met):

1. The CSpec is in screen mode (meta.type: screen)
2. The design was generated and executed in Figma (snapshot exists)
3. Total corrections during the fix cycle <= 2

If eligible:

This design qualifies for recipe extraction.
Creating recipe: {archetype} (from {name})

Extract recipe:

1. Take the final scene graph JSON (post-corrections)
2. Templatize: replace concrete text content with {{ param }} placeholders
3. Replace component keys with @lookup:ComponentName references
4. Define parameters based on variable content (title, items, section count, etc.)
5. Set initial confidence: 0.70 + (0.05 * 1) = 0.75 (first success)
6. Write to recipes/r-{archetype}-{nnn}.json
7. Update recipes/_index.json with the new recipe metadata