wayfinder
Pass
Audited by Gen Agent Trust Hub on Nov 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill instructions and associated Python scripts do not exhibit any malicious patterns. All operations are local, performing structural and logical validation of project management data in JSON format. The scripts decision_map_validator.py and tracer_slice_checker.py use standard Python libraries to process information and print reports to the console, without system-altering capabilities or network access.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied JSON files, which creates an ingestion point for untrusted data. However, this is a low-risk surface as the scripts are purely analytical. (1) Ingestion points: The validation scripts read project map data from local JSON files provided via the --input argument. (2) Boundary markers: No specific delimiters are used in the scripts' output to isolate processed data from the agent's prompt context. (3) Capability inventory: The scripts are limited to reading local files and printing validation results; they lack network, file-writing, or shell execution capabilities. (4) Sanitization: Input data is validated through standard JSON parsing and schema-specific checks for required fields.
Audit Metadata