skills/peterbamuhigire/chwezi-sdlc-documentation/17-ai-agent-action-catalogue-spec/Gen Agent Trust Hub
17-ai-agent-action-catalogue-spec
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data sources to generate security-critical configurations for tool usage.
- Ingestion points: The workflow requires reading multiple external files: AI_Agent_Feature_PRD_Spec.md, AI_Architecture_Spec.md, Multi_Tenancy_Architecture_Spec.md, and various API specifications as defined in logic.prompt and SKILL.md.
- Boundary markers: No explicit delimiters or instructions to ignore embedded commands are specified for the processing of these input files.
- Capability inventory: The skill is capable of defining and writing the Action_Catalogue_Spec.md and associated YAML tool definitions, which determine the autonomy levels, human-approval requirements, and kill-switch behaviors for an agent.
- Sanitization: There is no mention of sanitizing or validating the content of the input specifications. Maliciously crafted documentation could attempt to influence the agent to misclassify irreversible or high-risk tools as idempotent, potentially bypassing intended human-in-the-loop controls.
Audit Metadata