understand
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and traces repository code (Step 1), which acts as an untrusted data source. While the instructions do not specify explicit boundary markers or sanitization for code excerpts, the skill is constrained to produce code-grounded, plain-language explanations. The skill possesses file-writing capabilities and uses external research tools, but these are scoped to the intended documentation purpose.
- [DYNAMIC_EXECUTION]: The skill generates local HTML documentation files within the .skills/understand/ directory. It proactively addresses potential risks associated with web content execution by mandating that all pages be self-contained and strictly forbidding the use of remote scripts, external stylesheets, or CDNs.
Audit Metadata