scout
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill fetches and processes up to 10 external web pages to identify production archetypes, which creates a surface for indirect instructions hidden in technical content to manipulate the agent.
- Ingestion points: Full body content of web pages retrieved via dynamic research subagents.
- Boundary markers: The protocol does not specify distinct delimiters to isolate untrusted fetched text from the agent's internal instructions, although it requires a structured final report format.
- Capability inventory: The skill is capable of providing synthesized reports to the user and writing detailed dossiers to the local filesystem at '.agents/scratch/scout-dossier.md'.
- Sanitization: No explicit logic is provided for sanitizing or filtering instructions that may be embedded within the retrieved technical documentation.
Audit Metadata