titan-cu-bulletproof-v2listed

# Titan: cu-bulletproof-v2 ## Mandate Push Computer Use (CU) readiness from cu-titan v1's 88/100 to ≥95/100 by closing four deferred items: 1. **Safety-class enforcer wiring** — connect `claws_consent_prompt` + `claws_action_log_append` into every high-risk CU primitive (mouse_drag, key_combo across modifier keys, window_focus on unrecognized apps). 2. **Pattern induction → memory bridge** — when a CU workflow succeeds N times, auto-induce a `cu-pattern_*.md` memory via `claws_remember`. Bridge module at `extension/src/computer-use/pattern-induction.ts`. 3. **Tier 2-4 simulation coverage** — extend `.claude/packs/cu-simulations/` from current Tier 1 dominance to 8+ Tier 2 (cross-app), 6+ Tier 3 (long-horizon), and 3+ Tier 4 (adversarial) simulations. 4. **PPM signing infrastructure** — sign per-procedural-memory artifacts with HMAC; verification gate in `lib/ppm/verify.js`. Tamper detection on `.claude/packs/procedures/**`. Per [[root-cause-not-patch]] — fix architectural gaps, not symptoms. Audit cu-titan v1's `.local/blueprints/v083-cu-titan/READY-REPORT.md` deferred-items section before authoring. ## Deliverables | Artifact | Path | |---|---| | Safety enforcer wiring | `extension/src/computer-use/safety-enforcer.ts` (NEW or EXTEND) | | Pattern induction module | `extension/src/computer-use/pattern-induction.ts` (NEW) | | Memory bridge | `extension/src/computer-use/memory-bridge.ts` (NEW) | | Tier 2-4 simulations | `.claude/packs/cu-simulations/tier-{2,3,4}-*.md` (≥17