Security And Privacy Review

Choose one primary mode and read only its workflow. Add modes only when the same change crosses distinct security, AI-authority, or privacy boundaries.

Mode	Use when	Workflow
threat-modeling	Analyze assets, actors, entry points, trust boundaries, attack paths, controls, or residual risk.	Threat modeling
prompt-injection-defense	Review untrusted content influencing tools, retrieval, hidden instructions, external calls, or writes.	Prompt-injection defense
privacy-review	Review collection, consent, minimization, retention, deletion, exports, subprocessors, or user rights.	Privacy review

Do not load sibling workflows or artifacts prospectively. Stop before unauthorized exploitation, personal-data disclosure, production mutation, or legal/compliance conclusions reserved for owners.