explaining-codebases

Pass

Audited by Gen Agent Trust Hub on Jul 10, 2026

Risk Level: SAFEPROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8).
  • Ingestion points: In SKILL.md and references/code-intake.md, the agent is instructed to navigate repositories and read source files to identify entry points, module structure, and pull code snippets.
  • Boundary markers: The instructions do not define delimiters or provide warnings to the agent to ignore instructions potentially embedded within the analyzed codebase comments or files.
  • Capability inventory: The skill utilizes file-reading and directory-navigation capabilities to gather information.
  • Sanitization: No sanitization or validation protocols are described for handling extracted content before it is processed or incorporated into output.
  • [REMOTE_CODE_EXECUTION]: The skill involves dynamic code generation based on internal templates (Category 10).
  • Evidence: The file references/code-figure-archetypes.md provides JavaScript templates (IIFEs) used to create interactive figures on an HTML canvas.
  • Context: This generated code is intended for inclusion in the final interactive explainer output and is not designed for execution in the agent's local environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 10, 2026, 12:17 PM
Security Audit — agent-trust-hub — explaining-codebases