professional-brain

Pass

Audited by Gen Agent Trust Hub on Jun 23, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local Python scripts, brain_query.py and brain_write.py, to manage knowledge storage. These scripts use the standard library for searching and appending to files within the project's directory.\n- [PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by storing external content in a local 'brain' that is later used to guide agent responses.\n
  • Ingestion points: Data from various sources is ingested into the brain/ directory via the ingest command.\n
  • Boundary markers: Ingested data is stored without explicit markers to isolate potentially malicious instructions from factual content.\n
  • Capability inventory: The agent can read and write to the local filesystem using the provided scripts.\n
  • Sanitization: The skill does not perform sanitization of the ingested content before it is written to the markdown files.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 23, 2026, 12:52 PM
Security Audit — agent-trust-hub — professional-brain