A Practice Loop is a repeatable, AI-supported nursing workflow with a job, a standard, and a stopping rule. Instead of one-off prompting, each loop drives Claude through six pillars —
Trigger → Task → Standard → Verification → Iteration → Human Sign-Off
— printing a 10-point verification score, self-correcting anything below 8/10, halting on risk, and writing an audit trail to disk.
AI supports the workflow; the registered professional owns the judgement.
It maps directly onto the nursing process you already use: Assess → Plan → Intervene → Evaluate → Adjust.
| Skill | Turns this… | …into a DRAFT |
|---|---|---|
🎓 placement-support |
placement meeting notes | SMART action plan separating learning needs from conduct concerns |
🌱 preceptorship |
preceptee progress logs | 3-month review: confidence map + evidence gaps + reflective prompts |
💬 clinical-supervision |
supervision notes | restorative follow-up record: themes, actions, prompts |
⚖️ edi-intelligence |
aggregate workforce metrics | equity briefing using rate-based fair comparison |
📚 teaching |
a topic + audience level | session resource mapped to NMC proficiencies, with inclusive adjustments |
✅ action-tracking |
a meeting transcript | governance-ready action log; risks flagged for human grading |
🔄 revalidation |
revalidation evidence | portfolio-readiness summary against the NMC requirements, with gaps |
✍️ reflective-practice |
a practice experience | structured reflective-account scaffold mapped to the Code (CPD/revalidation) |
🔎 incident-reflection |
an anonymised incident/near-miss | just-culture learning summary (contributing factors + learning, not blame) |
♿ reasonable-adjustments-passport |
disclosed needs | Equality Act reasonable-adjustments passport (proposed, to agree) |
📄 policy-to-practice |
a national report/policy | plain-English summary + candidate local governance actions |
🧭 practice-loop-method |
— | the shared method that anchors all of the above |
Each loop ships with an anonymised example and a sample audit log in examples/.
In Claude Code:
/plugin marketplace add Clinical-Quality-Artifical-Intelligence/practice-loops
/plugin install practice-loops
Then invoke a loop by name — e.g. /practice-loops:placement-support — or in natural language: “run a placement support loop on these notes.”
- Anonymised inputs only. Never paste identifiable patient or staff data — the loops halt and ask if they detect it. No AI tool may process identifiable data without your organisation's Information Governance (IG) approval.
- Every output is a DRAFT pending sign-off by a named registered professional. The loop never finalises.
- Never use a loop to decide pass/fail or fitness to practise, diagnose or treat, determine mental capacity, or make final safeguarding, disciplinary, or employment decisions.
- This is not a medical device and does not replace professional judgement. The audit log records what the assistant did; it is not an independent assurance check.
A DCB0129 clinical safety case and hazard log for the plugin are in docs/safety/ (manufacturer drafts, pending Clinical Safety Officer sign-off).
claude --plugin-dir ./plugins/practice-loops # load without installing
# then /reload-plugins after edits
python3 scripts/validate.py # structural test harness
claude plugin validate ./plugins/practice-loops # official validatorAdding a loop? See CONTRIBUTING.md.
The full framework — the master guide, playbook, copy-paste templates, the infographic, and the source decks — lives in docs/framework/. The design spec and implementation plan are in docs/superpowers/.
Apache-2.0 · © 2026 Clinical Quality Artificial Intelligence · Practice Loops framework by Lincoln Gombedza
Technology teams build platforms, but nurses understand practice.