The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to check for and install updates from the author's GitHub repository.
Evidence: Found in SKILL.md under the 'UPDATE FIRST' section and in references/KNOWLEDGE.md.
Description: The agent is directed to fetch a version manifest from raw.githubusercontent.com/nha-in/nhcx-package/main/skills/manifest.json and download a zip archive from codeload.github.com/nha-in/nhcx-package/zip/refs/heads/main to replace the skill's own files.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted FHIR data from external hospitals, creating a potential surface for indirect prompt injection.
Ingestion points: Inbound HTTP routes /in/<path> and /v1/<path> as described in gateway/G8-receive.md.
Boundary markers: The skill relies on structured FHIR parsing and protocol header validation rather than explicit natural language delimiters.
Capability inventory: The skill is capable of database operations (database/), file system writes for message ledgers and archiving (gateway/G9-ledger.md, references/SCAFFOLDING.md), and outbound network communication with NHCX/ABDM services (gateway/G7-send.md, gateway/G10-beneficiary-registry.md).
Sanitization: Data is processed through decryption (gateway/G6-encryption.md) and structured parsing into FHIR resources (fhir/) before being utilized by the application logic.
[DYNAMIC_EXECUTION]: The skill generates and executes diagnostic scripts within the target repository.
Evidence: steps/LOG.md (LOGH section) and SKILL.md (STEPS section).
Description: The skill instructs the agent to create a report builder script (e.g., report.py) and shell wrappers in the target repository's nhcx-plan/ directory, then execute them at the end of every task to rebuild an HTML status report.

nhcx-payer-payment