11-ai-architecture-spec
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes multiple project documentation files ('HLD.md', 'Multi_Tenancy_Architecture_Spec.md', etc.) to generate architecture specifications. This creates an indirect prompt injection surface if those source files were to contain malicious instructions. The skill mitigates this through its workflow which requires tracing all statements back to approved inputs and mandates human review by architecture and security leads.
- Ingestion points: Multiple markdown files defined in 'logic.prompt'.
- Boundary markers: Workflow requires explicit traceability and observable acceptance conditions.
- Capability inventory: Markdown generation and document validation checks.
- Sanitization: Relies on human-in-the-loop review by specialized leads.
- [COMMAND_EXECUTION]: According to 'SKILL.md', the skill may execute document, syntax, or validation checks. These operations are limited to project deliverables and are intended for ensuring document quality and compliance with architectural standards.
- [SAFE]: The skill explicitly incorporates robust security architecture requirements, including the use of a model gateway as the sole egress point for provider calls, tenant identity propagation, content filtering, and sandboxing for tool execution.
Audit Metadata