Clinical Procedure Coding Assistant
Turn encounter documentation into claim-ready CPT and HCPCS procedure codes.
The skill requires abstracting performed services from clinical documentation and verifying HCPCS results; however, it does not explain sensitive-data handling, user confirmation, permission boundaries, data flows, recovery, rollback, or complete attribution. It also permits model-proposed CPT codes without a CPT connector. The main risks are visible but controls and traceability are incomplete, so 13 points are deducted.
The abstraction, lookup, and output flow is coherent, but SKILL.md requires validate_code while the README lists lookup_code, creating an interface inconsistency. There are no committed tests, abnormal-input handling, or diagnostic failure messages. Static evidence cannot justify more than 10 points; 7 reflects a plausible but fragile happy path, with 3 points deducted.
Invocation phrases, audience scenario, coverage, and exclusions are reasonably clear. Finer non-fit boundaries, regional or Chinese-language support, and reachability in mainland-China networks are not documented; core HCPCS functionality depends on external connectors. The scenario is clear but environment fit and trigger precision are under-evidenced, so 6 points are deducted.
The documentation has stepwise organization, connector notes, scope, and limitations, providing basic progressive disclosure. License metadata is unknown, and there is no versioning, changelog, maintenance owner, or update path; connector naming is also inconsistent. It is readable and usable but governance and maintenance documentation are incomplete, so 9 points are deducted.
The skill addresses the core task of deriving CPT and HCPCS candidates from encounter documentation and specifies one-code-per-line output. However, it excludes modifiers, NCCI edits, MUE units, and payer substitutions, limiting whether the result is truly claim-ready; CPT codes may also require licensed-source verification. Under static calibration the score cannot exceed 7; 5 reflects incomplete direct usability, with 10 points deducted.
The files define lookup requirements and some auditable coding rules, but contain no committed tests, CI coverage, representative outputs, or third-party execution evidence. Static review also cannot reproduce behavior. Evidence is therefore limited to partial auditability, warranting 3 points and a 7-point deduction.
- This is a static source review; the skill, connectors, and code lookups were not executed, so accuracy, availability, and failure behavior are unverified.
- Clinical documentation may contain sensitive health information, but the files specify no minimization, de-identification, storage, or transmission controls.
- The claim-ready framing is limited by the explicit exclusion of modifiers, NCCI edits, MUE units, and payer rules; qualified coding review and an authorized current code set remain necessary.
- Without a CPT connector, the workflow relies on model memory, and the README and SKILL.md do not fully agree on the HCPCS interface names.
What it does & when to use it
This skill is designed for professional coders who convert encounter notes and operative reports into procedure claims. It abstracts services actually performed, including evaluation and management, laboratory, imaging, medicine, surgery, supplies, devices, and facility-administered drugs. It then uses code-search and validation tools to identify CPT and HCPCS Level II codes while excluding planned services, prior results, bundled services, and unsupported reporting codes. It fits documentation-driven medical billing workflows with code lookup tools, but the source provides no standalone test suite or accuracy results.
Reads encounter notes, procedure narratives, laboratory results, medication-administration records, and supply or implant information; abstracts each distinct billable service performed; identifies CPT and HCPCS Level II codes; searches and validates codes with search_codes, validate_code, cpt_search_codes, and cpt_lookup_code when available; and outputs one claim-appropriate code per line while omitting unsupported matches, J#### codes, Q#### codes, and excluded quality-reporting codes.
- A professional coder converts an office, emergency, observation, or inpatient encounter note into procedure codes.
- A hospital coder reviews an interventional, endoscopic, or implant procedure narrative for separately reportable devices and facility drugs.
- A billing team adds venipuncture, laboratory, ECG, pulse-ox, injection, infusion, or vaccination codes documented during a visit.
- A coding reviewer separates performed services from planned services and identifies items included in a procedure's global package.
- A healthcare organization evaluates whether structured quality-measure or preventive-service documentation supports a reportable code.
Pros & cons
- Covers both CPT and HCPCS Level II, including physician services, devices, drugs, supplies, and non-physician services.
- Focuses on services documented as performed rather than services merely planned.
- Defines when to exclude quality-reporting codes, Category II tracking codes, and non-covered items.
- Supports encounter-date-based HCPCS searching and validation.
- CPT descriptors are AMA-licensed; without a CPT lookup connector, proposed CPT codes remain pending verification.
- It depends on available code-search, validation, or MCP tools, and the source does not document their standalone setup.
- The source provides no test suite, accuracy benchmark, or platform-specific test results.
- Its scope is procedure coding; capabilities of the other ten repository skills should not be attributed to it.
How to install
The skill is located at plugins/healthcare/skills/procedure-coding/ in the healthcare repository. To install the healthcare collection, run: /plugin marketplace add anthropics/healthcare, followed by /plugin install healthcare@healthcare. The source does not document a standalone installation method for this skill.
How to use
Provide an encounter note or operative report and use a trigger such as “code this encounter for procedures,” “what CPT codes apply,” “assign HCPCS codes,” or “code this op note.” The skill searches and validates available codes, then returns the applicable codes one per line.