Turn quality events into evidence-bound CAPA investigations with tested causes, linked actions, pre-registered effectiveness checks, and honest closure verdicts.
CAPA records become unreliable when a familiar diagram substitutes for evidence, training becomes the default fix, or a clean closure date matters more than proving the failure pathway changed. This skill keeps the investigation traceable from the trigger and source ledger through scope, cause tests, actions, effectiveness evidence, residual risk, and approval.
It builds a neutral problem statement, extent assessment, dated timeline, cause-hypothesis ledger, action-to-cause matrix, effectiveness protocol, open-decision log, and closure-readiness verdict. A bundled read-only checker catches broken evidence references, unsupported cause status, unlinked corrective actions, missing effectiveness definitions, invalid dates, and closure records that still contain open controlling items.
Use it for recurring nonconformities, complaints, audit findings, supplier issues, deviations, process failures, or a CAPA record that needs a skeptical review. It does not decide reportability, product disposition, patient causality, regulatory compliance, residual-risk acceptance, or CAPA closure. Those decisions stay with the authorized Quality, Regulatory, Safety, Legal, or management owner.
Input
Three seal failures were found across 18,000 units from Line 2 over six weeks. Two occurred after a tooling change; the third unit's build date is disputed. The supervisor says the operator skipped a check, but the batch record has no field for that check. Containment covers the current lot only. Build the CAPA investigation and say what can be approved now.
Output
A needs-evidence verdict; source and timeline reconciliation for the disputed unit; confirmed and potential scope; immediate containment gap; occurrence, escape, and system-level hypotheses; discriminating tests for tooling, process control, and record design; no unsupported operator-error conclusion; and an owned action and effectiveness protocol only where the cause evidence supports it.
Provide the event or CAPA record, controlled procedure and revision, applicable jurisdiction, product and process scope, available source records, current containment, decision owners, and known deadlines. The skill returns a traceable investigation record and the narrowest supportable verdict without filling missing requirements, causes, thresholds, tests, dates, or approvals with guesses.
The controlled CAPA or nonconformity procedure and revision, applicable jurisdiction, decision authority, risk method, and any contractual or product-specific requirements. Missing context remains an explicit gap.
Originating complaint, nonconformity, audit, deviation, trend, supplier, service, process, validation, change, training, or interview records available for the defined scope.
Investigation owner, Quality approver, functional owners, due dates, escalation paths, and the people authorized to decide reportability, field action, residual risk, and closure.
Required only for the optional read-only CAPA JSON checker. It uses the standard library and writes findings to stdout.
SKILL.md; CAPA evidence and closure rubric; reusable investigation, cause, action, effectiveness, and closure templates; read-only structured CAPA checker; Agent Skills interface metadata
No reviews yet.
Input
A customer reported intermittent loss of an alarm after a firmware update. Engineering reproduced it once in 40 attempts. Support has four similar tickets classified under different categories. The draft CAPA says risk is low, retraining is complete, and closure is due Friday. No Regulatory or Safety decision is attached.
Output
A needs-immediate-escalation verdict; linked complaint and ticket extent; the unsupported risk statement removed; containment, correction, and reportability referrals separated; reproducibility limits and detection bias recorded; cause hypotheses tied to firmware and classification evidence; and closure blocked until authorized safety, reporting, action, and effectiveness decisions are evidenced.
Input
A supplier corrective action changed an incoming-inspection step. The owner wants to close after 30 days because no defects were found, but only two lots arrived and the inspection sampling plan also changed. The procedure says the effectiveness window must be justified from exposure and risk. Review the record.
Output
An in-effectiveness-monitoring or needs-evidence verdict; implementation separated from effectiveness; the two-lot exposure and changed detection method treated as limitations; comparable baseline, denominator, opportunity, and sampling definitions requested; a sourced acceptance criterion and window rationale; explicit pass, fail, inconclusive, and extension rules; and no invented 30-, 60-, or 90-day default.
Trebuh
Updated Jul 13, 2026
Automated scan passed. It found 1 finding; highest severity info.
Scanned Jul 14, 2026.
Automated scanning reduces risk but cannot guarantee a package is safe for every environment.
Creator
Rravino
