investigate-capa-root-cause
À propos
Cette compétence guide les développeurs à travers une analyse structurée des causes racines et la gestion des actions correctives et préventives (CAPA) pour les écarts de conformité. Elle propose la sélection de méthodes (5-Pourquoi, diagramme d'Ishikawa), la conception d'actions et la vérification de leur efficacité. Utilisez-la pour traiter les observations d'audit, les écarts système ou les problèmes récurrents nécessitant une investigation systématique.
Installation rapide
Claude Code
Recommandénpx skills add pjt222/agent-almanac -a claude-code/plugin add https://github.com/pjt222/agent-almanacgit clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/investigate-capa-root-causeCopiez et collez cette commande dans Claude Code pour installer cette compétence
Documentation
Investigate CAPA Root Cause
Structured RCA + effective corrective/preventive actions for compliance deviations.
Use When
- Audit finding needs CAPA
- Deviation / incident in validated sys
- Regulatory inspection observation needs formal response
- Data integrity anomaly needs investigation
- Recurring issues → systemic root
In
- Req: Description of deviation / finding / incident
- Req: Severity (critical, major, minor)
- Req: Evidence from audit / investigation
- Opt: Prior related CAPAs / investigations
- Opt: Relevant SOPs, validation docs, sys logs
- Opt: Interview notes
Do
Step 1: Initiate
# Root Cause Investigation
## Document ID: RCA-[CAPA-ID]
## CAPA Reference: CAPA-[YYYY]-[NNN]
### 1. Trigger
| Field | Value |
|-------|-------|
| Source | [Audit finding / Deviation / Inspection observation / Monitoring alert] |
| Reference | [Finding ID, deviation ID, or observation number] |
| System | [Affected system name and version] |
| Date discovered | [YYYY-MM-DD] |
| Severity | [Critical / Major / Minor] |
| Investigator | [Name, Title] |
| Investigation deadline | [Date — per severity: Critical 15 days, Major 30 days, Minor 60 days] |
### 2. Problem Statement
[Objective, factual description of what happened, what should have happened, and the gap between the two. No blame, no assumptions.]
### 3. Immediate Containment (if required)
| Action | Owner | Completed |
|--------|-------|-----------|
| [e.g., Restrict system access pending investigation] | [Name] | [Date] |
| [e.g., Quarantine affected batch records] | [Name] | [Date] |
| [e.g., Implement manual workaround] | [Name] | [Date] |
→ Investigation initiated w/ clear problem statement + containment w/in 24h for critical findings.
If err: Containment can't be implemented immediately → escalate QA Director + document risk of delayed containment.
Step 2: Select Method
Choose based on complexity:
### Investigation Method Selection
| Method | Best For | Complexity | Output |
|--------|----------|-----------|--------|
| **5-Why Analysis** | Single-cause problems, straightforward failures | Low | Linear cause chain |
| **Fishbone (Ishikawa)** | Multi-factor problems, process failures | Medium | Cause-and-effect diagram |
| **Fault Tree Analysis** | System failures, safety-critical events | High | Boolean logic tree |
**Selected method:** [5-Why / Fishbone / Fault Tree / Combination]
**Rationale:** [Why this method is appropriate for this problem]
→ Method matches complexity — no fault tree for simple procedural, no 5-Why for complex systemic.
If err: First method doesn't reach convincing root → apply 2nd. Convergence across methods strengthens.
Step 3: Conduct RCA
Opt A: 5-Why
### 5-Why Analysis
| Level | Question | Answer | Evidence |
|-------|----------|--------|----------|
| Why 1 | Why did [the problem] occur? | [Immediate cause] | [Evidence reference] |
| Why 2 | Why did [immediate cause] occur? | [Contributing factor] | [Evidence reference] |
| Why 3 | Why did [contributing factor] occur? | [Deeper cause] | [Evidence reference] |
| Why 4 | Why did [deeper cause] occur? | [Systemic cause] | [Evidence reference] |
| Why 5 | Why did [systemic cause] occur? | [Root cause] | [Evidence reference] |
**Root cause:** [Clear statement of the fundamental cause]
Opt B: Fishbone (Ishikawa)
### Fishbone Analysis
Analyse causes across six standard categories:
| Category | Potential Causes | Confirmed? | Evidence |
|----------|-----------------|------------|----------|
| **People** | Inadequate training, unfamiliarity with SOP, staffing shortage | [Y/N] | [Ref] |
| **Process** | SOP unclear, missing step, wrong sequence | [Y/N] | [Ref] |
| **Technology** | System misconfiguration, software bug, interface failure | [Y/N] | [Ref] |
| **Materials** | Incorrect input data, wrong version of reference document | [Y/N] | [Ref] |
| **Measurement** | Wrong metric, inadequate monitoring, missed threshold | [Y/N] | [Ref] |
| **Environment** | Organisational change, regulatory change, resource constraints | [Y/N] | [Ref] |
**Contributing causes:** [List confirmed causes]
**Root cause(s):** [The fundamental cause(s) — may be more than one]
Opt C: Fault Tree
### Fault Tree Analysis
**Top event:** [The undesired event]
Level 1 (OR gate — any of these could cause the top event):
├── [Cause A]
│ Level 2 (AND gate — both needed):
│ ├── [Sub-cause A1]
│ └── [Sub-cause A2]
├── [Cause B]
│ Level 2 (OR gate):
│ ├── [Sub-cause B1]
│ └── [Sub-cause B2]
└── [Cause C]
**Minimal cut sets:** [Smallest combinations of events that cause the top event]
**Root cause(s):** [Fundamental failures identified in the tree]
→ RCA reaches fundamental cause (not symptom) w/ evidence per step.
If err: Analysis only symptoms ("user made err") → push deeper. Ask: "Why could user make that err? What control should've prevented?"
Step 4: Design Corrective + Preventive Actions
Distinguish correction vs corrective vs preventive:
### CAPA Plan
| Category | Definition | Action | Owner | Deadline |
|----------|-----------|--------|-------|----------|
| **Correction** | Fix the immediate problem | [e.g., Re-enable audit trail for batch module] | [Name] | [Date] |
| **Corrective Action** | Eliminate the root cause | [e.g., Remove admin ability to disable audit trail; require change control for all audit trail configuration changes] | [Name] | [Date] |
| **Preventive Action** | Prevent recurrence in other areas | [e.g., Audit all systems for audit trail disable capability; add monitoring alert for audit trail configuration changes] | [Name] | [Date] |
### CAPA Details
**CAPA-[YYYY]-[NNN]-CA1: [Corrective Action Title]**
- **Root cause addressed:** [Specific root cause from Step 3]
- **Action description:** [Detailed description of what will be done]
- **Success criteria:** [Measurable outcome that proves the action worked]
- **Verification method:** [How effectiveness will be checked]
- **Verification date:** [When effectiveness will be verified — typically 3-6 months after implementation]
**CAPA-[YYYY]-[NNN]-PA1: [Preventive Action Title]**
- **Risk addressed:** [What recurrence or spread this prevents]
- **Action description:** [Detailed description]
- **Success criteria:** [Measurable outcome]
- **Verification method:** [How effectiveness will be checked]
- **Verification date:** [Date]
→ Every action traces to specific root, has measurable success criteria, + effectiveness verification plan.
If err: Success criteria vague ("improve compliance") → rewrite specific + measurable ("zero audit trail config changes outside change control for 6 consecutive months").
Step 5: Verify Effectiveness
After implementation → verify actions worked:
### Effectiveness Verification
**CAPA-[YYYY]-[NNN] — Verification Record**
| CAPA Action | Verification Date | Method | Evidence | Result |
|-------------|------------------|--------|----------|--------|
| CA1: [Action] | [Date] | [Method: audit, sampling, metric review] | [Evidence reference] | [Effective / Not Effective] |
| PA1: [Action] | [Date] | [Method] | [Evidence reference] | [Effective / Not Effective] |
### Effectiveness Criteria Check
- [ ] The original problem has not recurred since CAPA implementation
- [ ] The corrective action eliminated the root cause (evidence: [reference])
- [ ] The preventive action has been applied to similar systems/processes
- [ ] No new issues were introduced by the CAPA actions
### CAPA Closure
| Field | Value |
|-------|-------|
| Closure decision | [Closed — Effective / Closed — Not Effective / Extended] |
| Closed by | [Name, Title] |
| Closure date | [YYYY-MM-DD] |
| Next review | [If recurring, when to re-check] |
→ Verification demonstrates root eliminated, not just action completed.
If err: Verification shows CAPA not effective → reopen investigation + develop revised actions. Don't close ineffective CAPA.
Step 6: Analyse Trends
### CAPA Trend Analysis
| Period | Total CAPAs | By Source | Top 3 Root Cause Categories | Recurring? |
|--------|------------|-----------|---------------------------|------------|
| Q1 20XX | [N] | Audit: [n], Deviation: [n], Monitoring: [n] | [Cat1], [Cat2], [Cat3] | [Y/N] |
| Q2 20XX | [N] | Audit: [n], Deviation: [n], Monitoring: [n] | [Cat1], [Cat2], [Cat3] | [Y/N] |
### Systemic Issues
| Issue | Frequency | Systems Affected | Recommended Action |
|-------|-----------|-----------------|-------------------|
| [e.g., Training gaps] | [N occurrences in 12 months] | [Systems] | [Systemic programme improvement] |
→ Trend analysis IDs systemic issues individual CAPAs miss.
If err: Trending reveals recurring roots despite CAPAs → CAPAs treating symptoms. Escalate to mgmt for systemic intervention.
Check
- Investigation initiated w/in timeline (24h critical, 72h major)
- Problem statement factual, no blame
- Method appropriate for complexity
- RCA reaches fundamental cause (not symptoms)
- Every step supported by evidence
- CAPAs distinguish correction, corrective, preventive
- Each CAPA has measurable success criteria + verification plan
- Effectiveness verified w/ evidence before closure
- Trend analysis reviewed ≥ quarterly
Traps
- Stop at symptom: "User made err" ≠ root. Root = why sys/process allowed err.
- CAPA = retraining: Retraining addresses only 1 possible root (knowledge). Real root = sys design flaw / unclear SOP → retraining won't prevent.
- Close w/o verification: Completing action ≠ verifying effectiveness. Closed CAPA w/o verification = regulatory citation waiting.
- Blame-oriented: Focus on who made err vs what allowed err → undermines quality culture, discourages reporting.
- No trending: Individual CAPAs seem unrelated, trending reveals systemic issues (e.g., "training" roots across multi systems = broken training prog).
→
conduct-gxp-audit— audits → findings → CAPAsmonitor-data-integrity— monitoring detects anomalies → investigationsmanage-change-control— CAPA-driven changes go thru change controlprepare-inspection-readiness— open/overdue CAPAs top inspection targetsdesign-training-program— root = training → improve prog
Dépôt GitHub
Compétences associées
executing-plans
DesignUtilisez la compétence executing-plans lorsque vous disposez d'un plan de mise en œuvre complet à exécuter par lots contrôlés avec des points de contrôle de revue. Elle charge et examine le plan de manière critique, puis exécute les tâches par petits lots (3 tâches par défaut) tout en rapportant la progression entre chaque lot pour une revue par l'architecte. Cela garantit une mise en œuvre systématique avec des points de contrôle de qualité intégrés.
requesting-code-review
DesignCette compétence délègue un sous-agent réviseur de code pour analyser les modifications apportées au code par rapport aux exigences avant de poursuivre. Elle doit être utilisée après avoir terminé des tâches, implémenté des fonctionnalités majeures, ou avant une fusion vers la branche principale. La revue aide à détecter précocement les problèmes en comparant l'implémentation actuelle avec le plan initial.
connect-mcp-server
DesignCette compétence fournit un guide complet permettant aux développeurs de connecter des serveurs MCP à Claude Code via les transports HTTP, stdio ou SSE. Elle couvre l'installation, la configuration, l'authentification et la sécurité pour intégrer des services externes tels que GitHub, Notion et des API personnalisées. Utilisez-la lors de la configuration d'intégrations MCP, de la configuration d'outils externes ou du travail avec le Protocole de Contexte de Modèle de Claude.
web-cli-teleport
DesignCette compétence aide les développeurs à choisir entre les interfaces Web et CLI de Claude Code en fonction de l'analyse des tâches, puis permet une téléportation transparente des sessions entre ces environnements. Elle optimise le flux de travail en gérant l'état et le contexte de la session lors du passage entre le web, la CLI ou le mobile. Utilisez-la pour des projets complexes nécessitant différents outils à diverses étapes.
