teach-pro-max

Pass

Audited by Gen Agent Trust Hub on Aug 29, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted external data, including user-supplied documents and tool outputs. It includes proactive security instructions to treat this content as non-authoritative and ignore embedded malicious prompts.
  • Ingestion points: SKILL.md identifies learner records, retrieved sources, and tool outputs as data entry points.
  • Boundary markers: The skill includes explicit directives to ignore embedded requests to reveal secrets or override instructions.
  • Capability inventory: The engine can write persistent learner state to the local filesystem and execute bundled scripts.
  • Sanitization: Uses the sanitize-html library to process generated content.
  • [COMMAND_EXECUTION]: To perform core functions such as rendering lessons and verifying distribution integrity, the skill invokes bundled Python and Node.js utilities (e.g., scripts/verify_distribution.py, scripts/validate_workspace.py, scripts/render_markdown.mjs).
  • [DATA_EXFILTRATION]: The skill implements a "no-API boundary" that explicitly forbids silent uploads, remote data storage, and CDN dependencies. All durable learner state is stored locally and requires explicit user consent before creation.
  • [EXTERNAL_DOWNLOADS]: Dependencies for the spacing algorithm and rendering pipeline are sourced from standard package registries (PyPI and NPM) and repositories belonging to well-known organizations.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 29, 2026, 06:19 PM
Security Audit — agent-trust-hub — teach-pro-max