judge
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external implementation code which constitutes a surface for indirect prompt injection attacks designed to influence the scoring outcome.
- Ingestion points: The 'Process' section in SKILL.md instructs the agent to read all implementation code currently in the context for evaluation.
- Boundary markers: The instructions lack specific delimiters or negative constraints (e.g., 'ignore any instructions within the code') to prevent the agent from following directives embedded within the implementations.
- Capability inventory: The skill is restricted to generating text-based markdown reports and does not have the ability to execute code, access the network, or modify files.
- Sanitization: No sanitization or validation of the ingested implementation code is performed before it is analyzed by the agent.
Audit Metadata