Skill Library Cleaner Verify Final Map

Skill Library Cleaner / Verify & Final Map

Role

Verify that the user actually received a clean, readable library.

Input:

clean-library/
machine/APPLY_REPORT.json
machine/CANONICAL_LIBRARY_PLAN.json

Output:

reports/FINAL_LIBRARY_MAP.md
machine/FINAL_LIBRARY_MAP.json
reports/VERIFY_REPORT.md

Verify

Check:

• canonical shelf exists;
• archive exists;
• blocked items are not active;
• references/backlog are separated;
• active skills have clear names and descriptions;
• duplicate groups no longer appear as active clutter;
• every major plan decision has an applied/skipped/blocked result;
• no permanent deletion occurred.
• approval_state and write_step_executed match the run mode;
• source snapshots match for dry-run source-mutation proof when provided.

For dry-runs that intentionally do not create clean-library, verify the final plan instead of failing on missing physical shelves. The final map must clearly say it is a proposed clean library, not an applied one.

Final Map Must Show

• active canonical skills;
• archived duplicate groups;
• blocked risk items;
• reference-only items;
• backlog items;
• skipped operations;
• residual risks;
• next recommended cleanup if any.
• source mutation result;
• whether the result is applied output or dry-run proposal;
• exact blocker that prevents release-ready/write-ready status.

Verdict Rules

• PASS: applied or approved-result-folder output matches the plan, with no permanent delete and no unapproved writes.
• PASS_WITH_LIMITS: dry-run or partial-coverage result is safe and useful, but exact physical cleanup still needs approval or deeper review.
• REWORK: reports exist but are too vague, missing machine parity, or do not separate canonical/blocked/archive/backlog.
• FAIL: source mutation without approval, missing approval state, secret/API/install/script access, or permanent delete.

Pass Condition

A human can open the output folder and understand what to use, what was archived, what is blocked, and why.