nutritional-specialist

Pass

Audited by Gen Agent Trust Hub on Mar 31, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill invokes a local Python script scripts/preferences_manager.py using subprocess.run and through direct Python imports to manage persistent data. This execution is limited to the skill's own logic and the storage of user configuration.\n- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it collects unvalidated user input (such as food preferences, dislikes, and notes), stores it locally, and then interpolates these strings directly into the agent's instructions for future responses.\n
  • Ingestion points: ~/.claude/nutritional_preferences.json (via the preferences_manager.py script).\n
  • Boundary markers: Absent; user preferences are mixed into the natural language workflow without delimiters or instructions to ignore embedded commands.\n
  • Capability inventory: The skill can read and write to the local file system and execute Python code based on the instructions in SKILL.md.\n
  • Sanitization: Absent; the skill does not filter or escape stored user data before using it to influence the agent's behavior.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 31, 2026, 04:02 AM
Security Audit — agent-trust-hub — nutritional-specialist