The Agent Skills Directory

[SAFE]: The skill establishes a rigorous process for defining AI agent requirements, incorporating established standards such as ISO/IEC/IEEE 29148, NIST AI RMF, and the EU AI Act. It includes validation logic in logic.prompt to ensure that critical safety controls, such as budget caps and human approver roles, are present in the output specification.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external project documentation to generate its output. While this presents an indirect prompt injection surface, the risk is minimal as the skill is restricted to document generation and lacks sensitive capabilities such as network exfiltration or system-level command execution.
Ingestion points: Processes project-specific artifacts including AI_Agent_Strategy_Doc.md, AI_Feature_PRD_Spec.md, AI_Feature_Strategy_Doc.md, Multi_Tenancy_Architecture_Spec.md, and pricing/packaging specifications.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are defined in the logic.prompt file when handling these inputs.
Capability inventory: Limited to reading input files and creating the AI_Agent_Feature_PRD_Spec.md output file.
Sanitization: Input content is processed without explicit sanitization or schema validation rules prior to template interpolation.

16-ai-agent-feature-prd-spec