teach

Pass

Audited by Gen Agent Trust Hub on Aug 31, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill architecture is designed to ingest and process untrusted data from external resources and user-provided learning missions. This creates a surface where a malicious resource or prompt could influence the agent's logic during the generation of lessons or learning records.\n
  • Ingestion points: The agent is instructed to read content from RESOURCES.md (which contains external URLs) and incorporate this information into ./lessons/ and ./learning-records/.\n
  • Boundary markers: The instructions do not specify the use of delimiters or protective headers to separate external data from the agent's internal instructions when processing resources.\n
  • Capability inventory: The agent has permissions to write markdown and HTML files to the local workspace and is encouraged to execute CLI commands to open these files for the user.\n
  • Sanitization: No explicit sanitization or validation steps are defined for the content fetched from external sources before it is interpreted or written to the filesystem.\n- [COMMAND_EXECUTION]: The skill instructs the agent to open generated lesson files using a CLI command (e.g., open, xdg-open). While this is a standard system operation for opening local files, it involves the agent executing shell commands based on file paths it has generated.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 31, 2026, 06:31 AM
Security Audit — agent-trust-hub — teach