perform-csv-assessment
について
このClaude Skillは、GxP規制環境においてGAMP 5手法を用いたコンピュータシステムバリデーション(CSV)評価を自動化します。リスクアセスメント、試験計画(IQ/OQ/PQ)、トレーサビリティマトリクスの作成といった主要なバリデーション作業を処理します。新規システムの導入時、既存システムへの重要な変更時、定期的な再バリデーション時、または規制当局の査察時などにご利用ください。
クイックインストール
Claude Code
推奨npx 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/perform-csv-assessmentこのコマンドをClaude Codeにコピー&ペーストしてスキルをインストールします
ドキュメント
Perform CSV Assessment
CSV via GAMP 5 risk-based for regulated envs.
Use When
- New computerized system in GxP env
- Existing validated system has significant change
- Periodic revalidation required
- Regulatory inspection prep → gap analysis
In
- Required: System desc (name, purpose, vendor, version)
- Required: Intended use + regulatory context (GxP scope)
- Required: GAMP 5 software category (1-5)
- Optional: Existing URS
- Optional: Vendor docs (specs, release notes, SOPs)
- Optional: Prev validation docs
Do
Step 1: Determine GAMP 5 category
Classify:
| Category | Type | Example | Validation Effort |
|---|---|---|---|
| 1 | Infrastructure software | OS, firmware | Low — verify installation |
| 3 | Non-configured product | COTS as-is | Low-Medium — verify functionality |
| 4 | Configured product | LIMS with config | Medium-High — verify configuration |
| 5 | Custom application | Bespoke R/Shiny app | High — full lifecycle validation |
→ Category assigned, rationale documented.
If err: ambiguous → default to higher cat + document rationale.
Step 2: Write URS
Numbered requirements:
# User Requirements Specification
## System: [System Name] v[Version]
## Document ID: URS-[SYS]-[NNN]
### 1. Purpose
[Intended use statement]
### 2. Functional Requirements
| ID | Requirement | Priority | Source |
|----|-------------|----------|--------|
| URS-001 | System shall calculate BMI from height and weight inputs | Must | Regulatory SOP-xxx |
| URS-002 | System shall generate audit trail entries for all data changes | Must | 21 CFR 11.10(e) |
| URS-003 | System shall export results in PDF format | Should | User request |
### 3. Non-Functional Requirements
| ID | Requirement | Priority | Source |
|----|-------------|----------|--------|
| URS-010 | System shall respond within 3 seconds for standard queries | Should | Usability |
| URS-011 | System shall restrict access via role-based authentication | Must | 21 CFR 11.10(d) |
### 4. Data Integrity Requirements
[ALCOA+ requirements: Attributable, Legible, Contemporaneous, Original, Accurate]
### 5. Regulatory Requirements
[Specific 21 CFR Part 11, EU Annex 11, or other applicable requirements]
→ All reqs have unique IDs, priorities, traceable source.
If err: no clear source/priority → flag for stakeholder review.
Step 3: Risk assessment
GAMP 5 risk-based via FMEA:
# Risk Assessment
## Document ID: RA-[SYS]-[NNN]
| Req ID | Failure Mode | Severity (1-5) | Probability (1-5) | Detectability (1-5) | RPN | Risk Level | Mitigation |
|--------|-------------|----------------|-------------------|---------------------|-----|------------|------------|
| URS-001 | Incorrect BMI calculation | 4 | 2 | 1 | 8 | Low | OQ test case |
| URS-002 | Audit trail entries missing | 5 | 3 | 3 | 45 | High | IQ + OQ + monitoring |
| URS-011 | Unauthorized access | 5 | 2 | 2 | 20 | Medium | OQ test + periodic review |
RPN = Severity × Probability × Detectability.
| RPN Range | Risk Level | Testing Requirement |
|---|---|---|
| 1–12 | Low | Basic verification |
| 13–36 | Medium | Documented test case |
| 37+ | High | Full IQ/OQ/PQ with retest |
→ Every URS requirement has corresponding risk row.
If err: unassessed → escalate to validation lead before proceed.
Step 4: Validation plan
# Validation Plan
## Document ID: VP-[SYS]-[NNN]
### Scope
- System: [Name] v[Version]
- GAMP Category: [N]
- Validation approach: [Prospective / Retrospective / Concurrent]
### Qualification Stages
| Stage | Scope | Applies? | Rationale |
|-------|-------|----------|-----------|
| IQ | Installation correctness | Yes | Verify installation, dependencies, configuration |
| OQ | Operational requirements | Yes | Verify functional requirements from URS |
| PQ | Performance under real conditions | [Yes/No] | [Rationale] |
### Roles and Responsibilities
| Role | Name | Responsibility |
|------|------|---------------|
| Validation Lead | [Name] | Plan, coordinate, approve |
| Tester | [Name] | Execute test scripts |
| System Owner | [Name] | Approve for production use |
| QA | [Name] | Review and sign-off |
### Acceptance Criteria
- All critical test cases pass
- No unresolved critical or major deviations
- Traceability matrix complete
→ Plan approved by stakeholders before test execution.
If err: no approved plan → don't proceed to test execution.
Step 5: Test protocols (IQ/OQ/PQ)
Test scripts per stage:
# Operational Qualification Protocol
## Test Case: TC-OQ-001
## Traces to: URS-001
**Objective:** Verify BMI calculation accuracy
**Prerequisites:**
- System installed per IQ protocol
- Test data set prepared
**Test Steps:**
| Step | Action | Expected Result | Actual Result | Pass/Fail |
|------|--------|-----------------|---------------|-----------|
| 1 | Enter height=180cm, weight=75kg | BMI displayed as 23.15 | | |
| 2 | Enter height=160cm, weight=90kg | BMI displayed as 35.16 | | |
| 3 | Enter height=0, weight=75kg | Error message displayed | | |
**Tester:** _________ Date: _________
**Reviewer:** _________ Date: _________
→ Every medium/high-risk req has ≥1 test case.
If err: missing test cases → add before execution.
Step 6: Traceability matrix
RTM links req → risk → test:
# Traceability Matrix
## Document ID: TM-[SYS]-[NNN]
| URS ID | Requirement | Risk Level | Test Case(s) | Test Result | Status |
|--------|-------------|------------|--------------|-------------|--------|
| URS-001 | BMI calculation | Low | TC-OQ-001 | Pass | Verified |
| URS-002 | Audit trail | High | TC-IQ-003, TC-OQ-005 | Pass | Verified |
| URS-003 | PDF export | Low | TC-OQ-008 | Pass | Verified |
| URS-011 | Role-based access | Medium | TC-OQ-010, TC-OQ-011 | Pass | Verified |
→ 100% URS requirements in matrix w/ linked test results.
If err: req no linked test → flag as validation gap.
Step 7: Validation summary report
# Validation Summary Report
## Document ID: VSR-[SYS]-[NNN]
### 1. Executive Summary
[System name] v[version] has been validated in accordance with [VP document ID].
### 2. Validation Activities Performed
| Activity | Document ID | Status |
|----------|-------------|--------|
| User Requirements | URS-SYS-001 | Approved |
| Risk Assessment | RA-SYS-001 | Approved |
| Validation Plan | VP-SYS-001 | Approved |
| IQ Protocol/Report | IQ-SYS-001 | Executed — Pass |
| OQ Protocol/Report | OQ-SYS-001 | Executed — Pass |
| Traceability Matrix | TM-SYS-001 | Complete |
### 3. Deviations
| Dev ID | Description | Impact | Resolution |
|--------|-------------|--------|------------|
| DEV-001 | [Description] | [Impact assessment] | [Resolution and rationale] |
### 4. Conclusion
The system meets all user requirements as documented in [URS ID]. The validation is considered [Successful / Successful with conditions].
### 5. Approval
| Role | Name | Signature | Date |
|------|------|-----------|------|
| Validation Lead | | | |
| System Owner | | | |
| Quality Assurance | | | |
→ Report references all deliverables w/ clear pass/fail conclusion.
If err: deviations unresolved → "conditional" status w/ CAPA refs.
Check
- GAMP 5 category assigned w/ rationale
- URS w/ numbered reqs + priorities + traceable source
- Risk assessment covers every URS req
- Validation plan approved before test execution
- Test protocols have prerequisite, step, expected, signature fields
- Traceability matrix links req → risk → test results
- VSR documents activities, deviations, conclusion
- All docs have unique IDs + version control
Traps
- Over-validation: Cat 5 effort on Cat 3 software = waste. Match effort to risk
- Missing traceability: reqs not tracing to tests = invisible gaps
- Test before plan: tests executed before plan approved → invalid results
- Ignore non-functional: security, perf, data integrity often overlooked
- Static validation: one-time event. Changes need re-assessment
→
setup-gxp-r-project— project structure for validated R envswrite-validation-documentation— IQ/OQ/PQ protocol + report writingimplement-audit-trail— audit trail for electronic recordsvalidate-statistical-output— stat output verificationconduct-gxp-audit— auditing validated systems
GitHub リポジトリ
関連スキル
content-collections
メタこのスキルは、Content Collections(Markdown/MDXファイルを型安全なデータコレクションに変換するTypeScriptファーストのツール)の本番環境でテストされた設定を提供します。Zodバリデーションによる型安全性を実現し、ブログ、ドキュメントサイト、コンテンツ重視のVite + Reactアプリケーション構築時にご利用ください。Viteプラグインの設定、MDXコンパイルから、デプロイ最適化、スキーマバリデーションまで、すべてを網羅しています。
polymarket
メタこのスキルは、開発者がPolymarket予測市場プラットフォームを活用したアプリケーション構築を可能にします。API統合による取引や市場データの取得に加え、WebSocketを介したリアルタイムデータストリーミングにより、ライブ取引や市場活動を監視できます。取引戦略の実装や、ライブ市場更新を処理するツールの作成にご利用ください。
creating-opencode-plugins
メタこのスキルは、開発者がコマンド、ファイル、LSP操作など25種類以上のイベントタイプにフックするOpenCodeプラグインを作成することを支援します。JavaScript/TypeScriptモジュール向けに、プラグイン構造、イベントAPI仕様、および実装パターンを提供します。カスタムイベント駆動ロジックでOpenCode AIアシスタントのライフサイクルをインターセプト、監視、または拡張する必要がある場合にご利用ください。
sglang
メタSGLangは、高性能なLLMサービングフレームワークであり、RadixAttentionプレフィックスキャッシュを活用したJSON、正規表現、エージェントワークフロー向けの高速で構造化された生成を特長とします。特にプレフィックスが繰り返されるタスクにおいて、大幅に高速な推論を実現し、複雑な構造化出力やマルチターン対話に最適です。制約付きデコードが必要な場合や、広範なプレフィックス共有を伴うアプリケーションを構築する場合は、vLLMなどの代替案ではなくSGLangを選択してください。
