distill

Pass

Audited by Gen Agent Trust Hub on Jun 13, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the Bash tool to run git commands for repository discovery and maintenance. It identifies the root of the project to set the distillation scope and performs auto-commits to the ~/.origin repository. These actions are restricted to local repository management.
  • [DATA_EXFILTRATION]: The skill reads and writes data within the user's home directory (~/.origin). This is required for its function of managing wiki pages derived from source memories. No sensitive file access or network communication was detected.
  • [PROMPT_INJECTION]: The skill processes memory data from the distill tool, creating an indirect prompt injection surface. The agent synthesizes this untrusted data into wiki prose. Evidence: 1. Ingestion point: distill tool output in SKILL.md. 2. Boundary markers: Absent. 3. Capability inventory: Bash, create_page, and update_page tools. 4. Sanitization: The skill implements a manual 'coherence check' where the agent evaluates memory content before synthesis. This surface is mitigated by the specific synthesis task and manual verification.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 13, 2026, 07:51 AM
Security Audit — agent-trust-hub — distill