The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted FHIR data (including claims, pre-authorizations, and communications) received from external hospital participants.
Ingestion points: Incoming messages are processed through callbacks/C1-callback-door.md and specific handlers for callbacks C2 through C11.
Boundary markers: Absent. The skill instructions do not specify the use of delimiters or warnings to ignore embedded instructions within processed FHIR content.
Capability inventory: The skill is capable of performing network operations via gateway.send, writing to the local database, and archiving data to the filesystem.
Sanitization: While data is validated against schemas, no specific sanitization is described to prevent instruction extraction by the LLM.
[EXTERNAL_DOWNLOADS]: The skill implements a self-update routine (UPDATE FIRST in SKILL.md) that downloads and replaces its own instruction files from the vendor's repository at github.com/nha-in/nhcx-package.
[COMMAND_EXECUTION]: The skill generates and runs integration reports using shell scripts (make-report.sh, make-report.bat) and a report builder script (report.py or similar).
[DYNAMIC_EXECUTION]: During the discovery step (L1.2), the skill writes a custom report builder script in the target environment's native language and subsequently executes it to produce a dashboard.

nhcx-payer-full