Summary

When the curator LLM archives skills without verified consolidation evidence (e.g., hallucinated umbrella names, fallback/no-evidence pruning), active workflows break silently. This adds a post-archive safety guard that:

• Detects skills removed by the LLM review pass
• Checks for tool-call evidence of real consolidation (write_file/patch/create/edit via skill_manage on an umbrella skill)
• If NO consolidation evidence exists: moves skills back from .archive/ to the active skills directory (fail-closed behavior)
• Logs a warning and annotates the run summary with the rollback count

Problem

Issue #29912: In a real run, 10 operational skills were archived in one pass with consolidated_this_run=0 and all pruned entries sourced from "fallback (model named missing umbrella, no tool-call evidence)". This broke active workflows because the original skill names were removed from normal lookup.

Fix

Added a post-archive validation guard in _llm_pass() inside run_curator_review() (agent/curator.py). After the LLM review completes:

1. Compare before_names vs after_names to detect removed skills
2. Scan tool calls for skill_manage with write_file/patch/create/edit actions (consolidation evidence)
3. If no consolidation evidence found: restore archived skills from .archive/ to skills/
4. Annotate the run summary with the rollback count

Testing

• Unit test for the safety guard logic
• Integration test: verify skills are restored when no consolidation evidence exists
• Integration test: verify skills remain archived when consolidation evidence exists

Fixes #29912