ask-astra
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external 'relevant files or evidence' which could contain malicious instructions designed to influence the advisor's recommendation.
- Ingestion points: File SKILL.md specifies providing 'relevant files or evidence' to the advisor model.
- Boundary markers: The skill does not define specific delimiters or instructions to ignore embedded commands within the provided evidence.
- Capability inventory: While the advisor is restricted to being read-only and cannot edit files, the calling agent evaluates its output to continue tasks, creating a risk if the advice is based on poisoned input.
- Sanitization: No sanitization or validation of the provided evidence is described.
Audit Metadata