life-process-inbox

Pass

Audited by Gen Agent Trust Hub on Aug 19, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill operates entirely within a local markdown vault environment. Access to system paths is abstracted through a configuration object (e.g., layout.resources_dir), preventing unauthorized file system traversal outside the intended scope.\n- [SAFE]: The agent's role is strictly defined in ROLE.md as a 'secretary' (chief of staff). This role explicitly forbids unilateral changes to project states or product implementation work, acting as a functional constraint against autonomous destructive actions.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes external data (markdown notes) which may contain instructions targeting the LLM. \n
  • Ingestion points: Processing loop in SKILL.md for notes where processed: false.\n
  • Boundary markers: The skill uses a 'grant' system (explicit per-turn permission) and mandates confirmation for ambiguous bulk moves.\n
  • Capability inventory: File move, delete, and content update capabilities.\n
  • Sanitization: No explicit sanitization of note content is mentioned, but the human-in-the-loop confirmation requirement serves as a mitigation.\n- [COMMAND_EXECUTION]: While the skill can delete or move files, these are standard vault management tasks. The ROLE.md constraints ('wait for confirm on material state') ensure the user retains control over file system operations.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 19, 2026, 05:22 AM
Security Audit — agent-trust-hub — life-process-inbox