contextual-intelligence

Pass

Audited by Gen Agent Trust Hub on Mar 1, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill functions as a local analysis tool and its behavior is consistent with its stated purpose of project fingerprinting and recommendation.
  • [PROMPT_INJECTION]: The skill ingests untrusted data from the file system (project manifest files and directory structures), which represents a surface for indirect prompt injection. However, the architecture mitigates this risk by converting extracted features into binary vectors for similarity calculations rather than interpolating raw text into prompts.
  • Ingestion points: Project manifest files (e.g., package.json, Dockerfile) and directory scanning via ProjectFingerprinter.
  • Boundary markers: None specified in the documentation.
  • Capability inventory: Local file system read access; no network operations or shell execution capabilities were detected.
  • Sanitization: The skill performs feature extraction and vectorization, which prevents file content from being processed as executable instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 1, 2026, 11:55 PM