The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes clinical requirements from an external planning document, creating a potential ingestion surface for untrusted data.
Ingestion points: The skill reads from docs/planning/data-model-brief.md to inform its design process.
Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are used when reading the planning file.
Capability inventory: The skill has the capability to write a new specification file to docs/planning/fhir-data-model.md.
Sanitization: No explicit sanitization or validation of the input file content is mentioned.
Context: This behavior is the primary intended purpose of the skill and the risk is minimal as the output is restricted to a documentation file.
[SAFE]: All external URLs and resources belong to established healthcare and technology organizations, including HL7, LOINC (Regenstrief Institute), SNOMED International, and the National Library of Medicine (NLM).
[SAFE]: No obfuscation, credential exposure, or remote code execution patterns were detected. File system access is limited to the expected project directory paths for documentation and planning.

fhir-data-model-design