ontological-documentation
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill operates by analyzing external codebases across multiple languages, creating a vulnerability where malicious instructions embedded in source code comments or strings could be interpreted as agent commands.
- Ingestion points: Targeted codebase files (Python, JavaScript, TypeScript) and configuration metadata processed during the discovery and extraction phases in files like 'SKILL.md' and scripts like 'extract_concepts.py'.
- Boundary markers: The instructions lack explicit delimitation (e.g., XML tags or clear 'ignore instructions' directives) to prevent the agent from being influenced by instruction-like content within the analyzed files.
- Capability inventory: The skill is granted extensive file-system access ('Read', 'Write', 'Grep', 'Glob') and executes Python-based static analysis scripts, which creates a capability chain between untrusted data and system modifications.
- Sanitization: There is no implementation of content sanitization or valid token filtering to ensure that code-derived text is treated strictly as data rather than potential instructions.
Audit Metadata