teach
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes
subprocess.runinassets/tex-figs.pyto invoke Ghostscript (gs) for converting vector graphics (EPS/PS) into images. Additionally, the provided test suite uses shell commands to execute the audit scripts during validation. These operations are limited to local files within the teaching workspace. - [DYNAMIC_EXECUTION]: As described in
references/subagent-dispatch.md, the skill dispatches background tasks to subagents for parallel processing. These tasks include lesson fact-checking, source extraction from LaTeX files, and generating interactive prototypes. - [INDIRECT_PROMPT_INJECTION]: The skill features a vulnerability surface where it ingests user-provided Markdown notes to autonomously generate HTML lessons and mission statements. Although the
afk-prompt-template.mdincludes explicit safety instructions to prevent the agent from writing outside the workspace or obeying embedded commands, processing untrusted input for content generation is a known attack vector.
Audit Metadata