security-review

A security review for LLM prompts and shell execution maps code diffs to trust boundaries across LLM API, persistence, and system execution layers. It guides tiered analysis to surface vulnerabilities and generate a prioritized fix plan.

Boundary-aware threat modeling identifies security risks by mapping each code diff to its specific trust boundary. It evaluates crossings across LLM API, persistence, and system execution tiers to isolate vulnerabilities during code reviews.

Detect prompt injection risks by applying a tiered security review to code diffs handling LLM API boundaries. The review maps changes to trust boundaries and provides structured findings with clear remediation guidance before merging.

Yes, this security review works for code writing to disk and invoking subprocesses by evaluating Tier 2 persistence and Tier 3 system execution boundaries. It assesses file persistence and subprocess invocations to surface boundary-crossing risks.

The best way to structure security findings for external integrations is using a mapped diff workflow with tiered analysis. It generates a standardized report with conclusive findings or a NO ISSUES FOUND status.

You need a tiered security analysis for code changes when handling LLM prompts, model outputs, file persistence, or subprocess invocations. It maps diffs to risk boundaries and guides stepwise evaluation across LLM API, persistence, and system execution layers.