security-checklists

Security review depth is determined by identifying the trust boundaries a code change crosses and loading only the boundary-matched checklists—such as access control, injection, or secrets—relevant to that diff. This avoids a one-size-fits-all checklist and focuses the review on applicable risks.

Reviewing LLM-agent workflows requires loading dedicated llm-agent security modules that check prompt-injection risks, authorization boundaries, and exceptional conditions. These checklists apply during both spec-stage design analysis and implementation review to catch LLM-specific security failure modes.

Threat modeling checklists are needed for spec-stage security analysis when designing features that cross trust boundaries, such as file-upload endpoints or LLM prompt-building logic. They enable proactive design review to identify access control, injection, and supply chain risks before implementation begins.

Yes, boundary-specific security checklists support reviewing access control and SSRF vulnerabilities through dedicated reference modules for outbound requests and authorization. The orchestrator-driven loading mechanism applies only the checklist modules matching the trust boundaries present in your code diff.

The best way to review secrets and configuration security is using boundary-matched checklists that load specific modules for secrets and configuration trust boundaries. This approach checks for hardcoded credentials, misconfigurations, and repo-specific helper-bypass risks relevant to the code change.

Yes, this security review approach handles supply chain and path handling risks through dedicated boundary-matched checklist modules covering file handling, path traversal, and dependency supply chain. It applies scanner-aware delegation to ensure these risks are checked during both spec-stage and implementation review.