technical-quality-perspective
Pass
Audited by Gen Agent Trust Hub on Aug 14, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input such as code diffs, architecture documents, and requirements. It includes instructions to distinguish facts from inference and follow a strict output format, which helps mitigate potential injection risks from the processed data. This represents a standard attack surface for analysis-oriented skills.
- Ingestion points: Project materials (code, diffs, requirements, contracts) provided via the
promptduring runtime. - Boundary markers: The skill requires findings to be evidence-backed and instructs the agent to use a specific report contract (Summary, Facts, Evidence, etc.).
- Capability inventory: None. The skill only generates text-based reports and does not invoke shell commands, network requests, or file writes.
- Sanitization: The instructions emphasize labeling inference and gaps, avoiding invented implementation facts.
Audit Metadata