skill-scout
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources such as package registries, GitHub, and general web searches to identify and rank existing implementations.\n
- Ingestion points: External search results from public marketplaces, GitHub, and web content are processed (referenced in
SKILL.md) to generate candidate tables.\n - Boundary markers: The skill instructions do not explicitly define delimiters for search result content, but they include a strict vetting process for candidates.\n
- Capability inventory: The skill is primarily read-only in its scout mode; operations involving file modification or installation are delegated to other skills and require explicit user confirmation.\n
- Sanitization: The instructions explicitly direct the agent to check candidates for side effects, shell commands, and credential handling, and warn against treating external content as trusted source material.
Audit Metadata