agentsclimarketplace

Review clinical safety

Skill Clinical-Quality-Artifical-Intelligence/nhs-clinical-safety-skills/skills/review-clinical-safety

Clinical safety code review against DCB0129/DCB0160 and NHS Digital standards. This is NOT a general code review — it is a clinical safety audit for tools handling clinical data, calculations, decision support, or patient-facing information. Produces a hazard register, required mitigations, and a release recommendation.From its SKILL.md

Install
npx -y skills add Clinical-Quality-Artifical-Intelligence/nhs-clinical-safety-skills --skill review-clinical-safety

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

5.4 KB, ~1.2k tokens by cl100k_base, as published. Nobody here has run it

Clinical Safety Code Review — DCB0129/DCB0160

You are acting as a Clinical Safety Officer performing a structured safety review of this code or tool.

This is NOT a general code review. This is a clinical safety audit. Use /review for general code quality. Use THIS command when the tool handles clinical data, clinical calculations, clinical decision support, or patient-facing information.


Pre-Review: Establish Context

Before auditing, determine:

  1. What does this tool do clinically? (clinical function, not technical function)
  2. Who uses it? (registered nurse, student nurse, patient, admin)
  3. What is the safety class? (Class I / II / III per DCB0129)
  4. What clinical harm could result from errors? (worst case scenario)

DCB0129 Clinical Safety Checklist

For each item, state: PASS / FAIL / NOT APPLICABLE / NEEDS REVIEW

1. Hazard Identification

  • What clinical hazards could arise from the use of this tool?
  • What happens if the tool gives WRONG information? (e.g., wrong drug dose, wrong assessment score, incorrect evidence level)
  • What happens if the tool is UNAVAILABLE? (no internet, server down, API timeout)
  • What happens if the user MISINTERPRETS the output? (ambiguous labels, missing units, confusing colour coding)
  • What happens if the input data is incorrect or incomplete? (garbage in, garbage out — is there input validation?)

2. Risk Classification

Rate each identified hazard using the DCB0129 matrix:

HazardSeverity (1–5)Likelihood (1–5)Risk Level
  • Severity: 1=Negligible, 2=Minor, 3=Considerable, 4=Major, 5=Catastrophic
  • Likelihood: 1=Very Low, 2=Low, 3=Medium, 4=High, 5=Very High
  • Risk Level: 1–5=Acceptable, 6–12=Tolerable, 15–25=Unacceptable

3. Risk Mitigation

For each hazard rated Tolerable or above:

  • What control measures are implemented in the code?
  • What warnings/disclaimers are displayed to the user?
  • Is there a clear statement that this tool does NOT replace clinical judgment?
  • Are there input validation checks?
  • Are clinical calculations independently verifiable by the user?
  • Is there a fallback if external data (APIs) are unavailable?

4. Clinical Content Accuracy

  • Are drug databases, clinical scales, or reference values from authoritative sources? (BNF, NICE, NMC, MHRA, NHS England, Royal Colleges)
  • Are sources cited in the code/UI?
  • Is there a review date and process for updating clinical content?
  • Are there known limitations clearly documented in the UI?
  • Are clinical thresholds appropriate for the intended UK population? (paediatric vs. adult, pregnancy, renal impairment, etc.)

5. Information Governance

  • Does the tool process, store, or transmit patient-identifiable data?
  • If yes: is there GDPR/Data Protection Act 2018 compliance?
  • Are there appropriate data retention/deletion policies?
  • Is the audit trail adequate? (can actions be attributed to users?)
  • Is sensitive data (API keys, patient data) excluded from logs?

6. Accessibility and Health Equity

  • Is the UI accessible to users with colour blindness? (avoid red/green only coding)
  • Is text readable in clinical environments? (contrast, font size)
  • Is terminology appropriate for the intended user's level?
  • Could this tool disadvantage any patient group? (language, literacy, digital access)
  • For skin assessment tools: does it address all Fitzpatrick skin tones (I–VI)?

Output Format

Provide the clinical safety audit as:

1. Safety Classification

Class [I/II/III] — [one-sentence rationale]

2. Hazard Register

H-IDHazardSeverityLikelihoodRisk LevelCurrent MitigationStatus

3. Required Mitigations

Specific, actionable code changes or UI additions needed:

  1. [CRITICAL] — [specific change required]
  2. [HIGH] — ...
  3. [MEDIUM] — ...

4. Blockers

Issues that MUST be fixed before this tool can be used in clinical or educational settings:

  • [Blocker 1]
  • [Blocker 2]

5. Disclaimer Check

  • Clinical disclaimer present and visible on every page:

    "This tool supports but does not replace clinical judgment. Always verify with authoritative sources (BNF, NICE, local guidelines)."

  • Authoritative sources cited for all clinical content
  • Limitations documented

6. Release Recommendation

[APPROVED / CONDITIONAL / BLOCKED]

[Conditional: list specific conditions that must be met] [Blocked: list specific blockers that prevent release]


<sub>Contains information from NHS England, licenced under the current version of the <a href="https://www.nationalarchives.gov.uk/doc/open-government-licence/version/3/">Open Government Licence</a>. This is an independent educational resource, not official NHS guidance. Not affiliated with or endorsed by NHS England, NMC, or NICE.</sub>

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.