brain-taxonomist

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill requires the agent to execute shell commands using the gbrain utility and jq. It specifically mandates passing a source identifier (--source <id>) to these commands. This creates a potential command injection vector if the identifier is sourced from untrusted input and interpolated directly into the shell string without sanitization.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes arbitrary 'brain page' content to determine the primary subject and filing path. This analysis surface is susceptible to indirect prompt injection, where malicious text inside a page could attempt to influence the agent's classification logic or trigger unnecessary schema creation workflows.
  • Ingestion points: Content of new brain pages, bulk imports, and ambiguous text files provided as input.
  • Boundary markers: None are explicitly defined in the instructions to separate the data being analyzed from the agent's processing instructions.
  • Capability inventory: Execution of shell commands (gbrain, jq) and recommendation of filesystem paths.
  • Sanitization: No specific sanitization, filtering, or validation steps for the input content are described.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 04:38 PM
Security Audit — agent-trust-hub — brain-taxonomist