pathfinder

Pass

Audited by Gen Agent Trust Hub on Apr 25, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill acts as an architectural orchestrator that maps features and identifies logic duplication through static analysis of local repository files.
  • [SAFE]: No security concerns such as credential exposure, network exfiltration, or remote code downloads were identified within the skill instructions.
  • [SAFE]: The skill's operations are limited to reading repository source code and writing markdown documentation artifacts to a designated directory (PATHFINDER-<YYYY-MM-DD>/).
  • [SAFE]: Indirect prompt injection surface analysis: (1) Ingestion points: Reads local README, CLAUDE.md, and source code files. (2) Boundary markers: Enforces a 'Subagent Reporting Contract' requiring exact file paths and line ranges for all data. (3) Capability inventory: Restricted to repository file reads and writing documentation. (4) Sanitization: Relies on structured output formatting and human review of generated prompts before further action.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 25, 2026, 08:39 PM