Use this skill whenever building, modifying, or testing the CDI agent, generating physician queries, writing CDI prompts, or detecting documentation improvement opportunities. Triggers on: CDI, clinical documentation improvement, physician query, documentation gap, AHIMA, non-leading query, query compliance, creatinine rise, KDIGO, SIRS criteria, specificity upgrade opportunity, underdocumented condition, or any task involving asking physicians to clarify clinical documentation. Also use when editing src/agents/cdi_agent.py, src/prompts/cdi_query.py, or any code that generates queries sent to physicians. If the task mentions querying a doctor, improving documentation, or detecting underdocumented conditions — load this skill. Even if the user says "CDI" without elaboration, load this skill.
Use this skill whenever working on DRG calculation, revenue impact analysis, coding suggestion ranking, CFO-facing reports, or any component that computes or displays financial impact of coding changes. Triggers on: DRG, diagnosis related group, revenue impact, reimbursement, CC/MCC upgrade, DRG weight, DRG tier, base rate, case mix index, coding improvement value, CFO dashboard, revenue narrative, compliance review threshold, the $5000 guardrail (G-SOFT-003), or any dollar-amount calculation from coding changes. Also use when editing src/core/drg/, src/agents/drg_agent.py, src/prompts/drg_analysis.py, or any code that calculates or displays financial amounts. If the task involves money, revenue, or financial impact of coding — load this skill.
Use this skill whenever working on FHIR client code, EHR integration, clinical data retrieval, encounter/patient/observation resource parsing, authentication with Epic or Cerner, or any component that reads from or writes to a FHIR API. Triggers on: FHIR, HL7, Epic, Cerner, DocumentReference, Encounter, Patient, Observation, SMART on FHIR, OAuth token, rate limit, circuit breaker, DegradedResult from API calls, C-CDA, LOINC code, note extraction, encounter class mapping, observation-to-inpatient conversion, or HTTP 429/503 retry logic for clinical APIs. Also use when editing src/core/fhir/, src/mcp/fhir_tools.py, or any code that makes HTTP calls to an EHR system. If the task involves getting clinical data from an external system or parsing FHIR resources — load this skill.
MANDATORY for ANY code change in this healthcare project. Use this skill when writing, reviewing, or modifying ANY function that touches logging (structlog, print, log.*), error handling (try/except, raise, Error classes), API responses, middleware, audit trails, Pydantic models with patient fields, FHIR resource parsing, test fixtures containing clinical data, or .env/secrets configuration. If you are writing code in src/, tests/, or scripts/ — load this skill. If you see the words "log", "error", "patient", "encounter", "clinical", "note", "PHI", "HIPAA", "audit", "redact", or "filter" in the task — load this skill. When in doubt, load this skill. It is practically impossible to write code in this project without needing HIPAA guidance.
Session-end cleanup and security audit for the Healthcare Digital FTE project. Run this skill at the end of every Claude Code session before closing. Invoke whenever the user says "done", "closing", "end session", "wrap up", "hygiene", or "/hygiene". Also invoke proactively when the user says they're finished working for the day, finishing a task, or wrapping up a coding session. Performs: secret scan, claude.md length guard, prompt injection scan, dead file removal, test suite health check, PHI leak scan, dependency audit, ADR index compliance check, skill file validity check, and final session summary report.
Session-end cleanup and security audit for the Healthcare Digital FTE project. Run this skill at the end of every Claude Code session before closing. Invoke whenever the user says "done", "closing", "end session", "wrap up", "hygiene", or "/hygiene". Also invoke proactively when the user says they're finished working for the day, finishing a task, or wrapping up a coding session. Performs: secret scan, claude.md length guard, prompt injection scan, dead file removal, test suite health check, PHI leak scan, dependency audit, ADR index compliance check, skill file validity check, and final session summary report.
Use this skill whenever working on the coding agent, rules engine, CDI agent, DRG calculator, or ANY component that suggests, validates, sequences, or filters ICD-10 codes. Triggers on: ICD-10, ICD10, diagnosis code, coding guideline, Excludes 1, uncertain diagnosis, outpatient vs inpatient coding, POA indicator, combination code, Code First, Use Additional, manifestation code, specificity upgrade, CC/MCC status, or any reference to CMS coding rules. Also use when editing src/core/icd10/, src/agents/coding_agent.py, src/agents/cdi_agent.py, src/prompts/coding_extraction.py, tests/clinical/, or any file that imports from src.core.icd10. If the task involves clinical coding in ANY way, load this skill — even if the user does not explicitly mention ICD-10.
Use this skill whenever working on denial prevention, pre-submission claim validation, prior authorization checks, coding agent denial risk assessment, appeal letter generation, or any component that predicts or prevents insurance claim denials. Triggers on: denial, prior auth, PA, medical necessity, LCD, NCD, NCCI edit, timely filing, claim rejection, payer rules, unbundling, modifier, appeal, eligibility verification, filing deadline, documentation requirements for procedures, or any reference to insurance payer behavior. Also use when editing code related to claim submission, payer interaction, denial prediction, or appeal generation. If the task involves insurance, payers, claims, preventing rejections, or understanding why claims get denied — load this skill.