explaining-implementation
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data such as source code, pull requests, and algorithm descriptions to generate hierarchical explanations. This creates a surface for indirect prompt injection if the processed content contains malicious instructions intended to influence the agent's output behavior.
- Ingestion points: Processes repository files, diffs, and user-provided technical descriptions in SKILL.md under the 'Research' step.
- Boundary markers: The skill does not define explicit delimiters for ingested data, although it instructs the agent to 'read in silence' before starting the explanation walk.
- Capability inventory: The skill is restricted to reading provided content and generating markdown text. It lacks network access, write capabilities, or tool-invocation permissions.
- Sanitization: No explicit sanitization or filtering of instructions embedded within the source code is mentioned.
- [DATA_EXFILTRATION]: The skill explicitly mandates the use of local, relative file paths for all code references. It contains a specific instruction to 'Never substitute a hosted or remote view of the code', ensuring that code analysis remains within the user's local environment and preventing the redirection of the user to potentially malicious external mirrors.
Audit Metadata