paw-cra-setup

Pass

Audited by Gen Agent Trust Hub on Apr 2, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local Python scripts (merge-config.py, merge-help-csv.py) to manage project configuration and registry files. It also uses standard filesystem commands (mkdir -p) to create necessary project directories.
  • [DATA_EXFILTRATION]: The skill collects sensitive data including API keys for Fal.ai, ElevenLabs, and Pexels. These are stored in a specific user-level configuration file (.pawbytes/config/config.user.yaml) which the skill instructions specify should be gitignored, preventing accidental exposure in shared repositories.
  • [PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it ingests user-provided strings (e.g., user name, language preferences) and writes them into configuration files and a capability registry (module-help.csv). These files are subsequently used to influence agent behavior and menu displays.
  • Ingestion points: User responses to configuration prompts in SKILL.md.
  • Boundary markers: Not explicitly used in the target YAML/CSV files.
  • Capability inventory: Local command execution (python3, mkdir) and project-level file writing.
  • Sanitization: Employs yaml.safe_load() in its configuration scripts to prevent arbitrary code execution during YAML parsing.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 2, 2026, 09:36 AM