invariant-guard
Pass
Audited by Gen Agent Trust Hub on Oct 24, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides structured instructions for improving code quality and correctness through formal reasoning. It contains no executable scripts, network requests, or sensitive data access.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied algorithm descriptions to generate code and verification logic. This represents a standard interaction surface where the assistant ingests untrusted input to produce output. The skill's specific focus on pre-generation validation and formal contracts provides a structured reasoning framework for interpreting user requests.
- Ingestion points: User requests for algorithm implementation or code review (SKILL.md instructions).
- Boundary markers: The protocol defines specific sections (Contract, Invariants, etc.) for the assistant's response.
- Capability inventory: Code generation and logical analysis.
- Sanitization: Not explicitly defined; relies on the agent's core safety protocols.
Audit Metadata