nooa-middleware-hooks
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill establishes an extensibility framework that ingests untrusted data via middleware and observers in
SKILL.md. - Ingestion points: Untrusted content is processed through
LLMCallContext.messagesandexecute_pythoncontext objects inSKILL.md. - Boundary markers: No explicit boundary markers or isolation techniques for untrusted data are defined in the middleware logic provided in
SKILL.md. - Capability inventory: The middleware has capabilities to modify prompt content, block operations, and alter code execution parameters as described in
SKILL.md. - Sanitization: Although sanitization examples like secret redaction are provided, the framework provides raw access to message buffers without mandatory filtering.
Audit Metadata