archive-prdlisted

# PRD Archival System Archive completed stories to keep prd.json fast and small. Keep only last 3 sprints active. ## "archive" Command ``` 1. Read prd.json 2. Separate stories with isArchivable() from scripts/prd-states.js: - ARCHIVE: isArchivable(story) === true (passes===true, i.e. isDone) - KEEP: everything else — null, false, "deferred", "needs-setup", a MISSING passes key, any unrecognised value, and type="qa" (even passed QA stories stay, for re-testing) 3. PROVE THE SPLIT BEFORE WRITING: archive-count + keep-count must equal the before-count, and no story id may appear in both sets or in neither. If the invariant fails, STOP — write nothing, report the ids that fell through. This runs BEFORE step 4, not after: a loss detected after the write is a loss. 4. Create archive file: prd-archives/prd-archive-YYYY-MM.json (a TRACKED path) 5. Update main prd.json with summary 6. Report: "Archived X stories, Y remain active" with both counts and the total ``` **Why the two-bucket version was destructive:** the old split (ACTIVE: false|null|qa / COMPLETED: true) matched neither bucket for `"deferred"`, `"needs-setup"`, and keyless stories — they were written to neither file and silently deleted. `isArchivable()` exists because of exactly that incident (its own comment block records it); route through it rather than re-deriving buckets. ## New prd.json Schema (After Archive) ```json { "project": "Project Name", "version":