devtools-webmcp
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides a framework for registering tools that allow AI agents to inspect internal library state, which constitutes a potential surface for indirect prompt injection if the data being inspected contains untrusted instructions.\n
- Ingestion points: Data enters the agent's context through the
executefunction's return value inSKILL.md.\n - Boundary markers: The skill supports the use of
untrustedContentHintto inform the agent about potentially unsafe content.\n - Capability inventory: Registered tools are capable of reading and summarizing internal state (e.g., query caches).\n
- Sanitization: The implementation relies on the developer-defined
summarizefunction to handle data representation, with built-in support for security hints.
Audit Metadata