What problem does it solve? Changes that create, move, or delete trust boundaries often ship without structured threat analysis, leaving residual risks unnamed and unowned. This Skill runs a disciplined STRIDE threat-modeling pass over exactly one scoped subject and produces a classified, evidence-graded findings report without ever fixing code or authorizing anything. ## Core Features & Use Cases - Bound-subject analysis: Binds one diff, subsystem, or trust boundary as an immutable snapshot before any lens fires, refusing whole-repository scans. - Six STRIDE lenses: Runs spoofing, tampering, repudiation, information disclosure, denial of service, and elevation-of-privilege lenses with a per-element coverage vocabulary (applicable, not_applicable, deferred, out_of_scope). - Structured findings report: Emits seed-shaped findings with severity, STRIDE letters, evidence classes, acceptance criteria, and a human-only risk disposition column. - Use Case: Before accepting a change that touches credential handling, run the skill against that diff to enumerate spoofing and information-disclosure threats, then hand the report to a human for risk disposition. ## Quick Start Ask the agent to threat-model the current diff touching the authentication boundary using the sdlc-threat-model skill and return the findings report.
