teach
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions advise the agent to use CLI commands to open generated lesson files (e.g., HTML files in the ./lessons/ directory) for the user. This is a standard functional capability for a local development or teaching assistant and does not involve arbitrary or dangerous command construction.
- [INDIRECT_PROMPT_INJECTION]: The skill relies on reading and interpreting data from several workspace files that are subject to user input, which creates a surface for indirect instructions to influence the agent's behavior.
- Ingestion points: The agent reads the MISSION.md file, the RESOURCES.md file, and various learning records from the ./learning-records/ directory to determine the context and state of the teaching session.
- Boundary markers: The skill uses defined Markdown templates (e.g., MISSION-FORMAT.md) to structure the data, which acts as a rudimentary boundary between data and instructions.
- Capability inventory: The agent has the capability to write various file types (.html, .md) to the local filesystem, create directories, and execute shell commands to open files.
- Sanitization: There is no mention of explicit sanitization or filtering of the content ingested from the workspace files beyond the requirement to follow the specified Markdown formats.
Audit Metadata