setup-gxp-r-project
About
This skill sets up an R project structure compliant with GxP regulations like 21 CFR Part 11, including validated environments and documentation for change control. Use it to initiate R analyses in regulated pharmaceutical or clinical trial environments where electronic records must meet audit standards. It provides the foundational framework for implementing regulatory-compliant computing for submissions.
Quick Install
Claude Code
Recommendednpx 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/setup-gxp-r-projectCopy and paste this command in Claude Code to install this skill
Documentation
Set Up GxP R Project
R project structure → GxP reg validated computing reqs.
Use When
- R analysis in regulated env (pharma, biotech, med devices)
- R for clinical trial analysis
- Validated compute env for regulatory submissions
- Impl 21 CFR Part 11|EU Annex 11
In
- Required: Project scope + reg framework (FDA|EMA|both)
- Required: R ver + pkg vers to validate
- Required: Validation strategy (risk-based)
- Optional: Existing SOPs for computerized systems
- Optional: QMS integration reqs
Do
Step 1: Validated Project Structure
gxp-project/
├── R/ # Analysis scripts
│ ├── 01_data_import.R
│ ├── 02_data_processing.R
│ └── 03_analysis.R
├── validation/ # Validation documentation
│ ├── validation_plan.md # VP: scope, strategy, roles
│ ├── risk_assessment.md # Risk categorization
│ ├── iq/ # Installation Qualification
│ │ ├── iq_protocol.md
│ │ └── iq_report.md
│ ├── oq/ # Operational Qualification
│ │ ├── oq_protocol.md
│ │ └── oq_report.md
│ ├── pq/ # Performance Qualification
│ │ ├── pq_protocol.md
│ │ └── pq_report.md
│ └── traceability_matrix.md # Requirements to tests mapping
├── tests/ # Automated test suite
│ ├── testthat.R
│ └── testthat/
│ ├── test-data_import.R
│ └── test-analysis.R
├── data/ # Input data (controlled)
│ ├── raw/ # Immutable raw data
│ └── derived/ # Processed datasets
├── output/ # Analysis outputs
├── docs/ # Supporting documentation
│ ├── sop_references.md # Links to relevant SOPs
│ └── change_log.md # Manual change documentation
├── renv.lock # Locked dependencies
├── DESCRIPTION # Project metadata
├── .Rprofile # Session configuration
└── CLAUDE.md # AI assistant instructions
→ Complete dir structure exists w/ all listed dirs.
If err: missing → mkdir -p. Verify in correct project root. Existing → create only missing, don't overwrite.
Step 2: Validation Plan
validation/validation_plan.md:
# Validation Plan
## 1. Purpose
This plan defines the validation strategy for [Project Name] using R [version].
## 2. Scope
- R version: 4.5.0
- Packages: [list with versions]
- Analysis: [description]
- Regulatory framework: 21 CFR Part 11 / EU Annex 11
## 3. Risk Assessment Approach
Using GAMP 5 risk-based categories:
- Category 3: Non-configured products (R base)
- Category 4: Configured products (R packages with default settings)
- Category 5: Custom applications (custom R scripts)
## 4. Validation Activities
| Activity | Category 3 | Category 4 | Category 5 |
|----------|-----------|-----------|-----------|
| IQ | Required | Required | Required |
| OQ | Reduced | Standard | Enhanced |
| PQ | N/A | Standard | Enhanced |
## 5. Roles and Responsibilities
- Validation Lead: [Name]
- Developer: [Name]
- QA Reviewer: [Name]
- Approver: [Name]
## 6. Acceptance Criteria
All tests must pass with documented evidence.
→ Plan complete w/ scope, GAMP 5 risk categories, validation activities matrix, roles, acceptance criteria. References specific R ver + reg framework.
If err: framework unclear → consult QA dept for SOPs. Don't proceed validation activities until plan reviewed+approved.
Step 3: Lock Deps w/ renv
# Initialize renv with exact versions
renv::init()
# Install specific validated versions
renv::install("[email protected]")
renv::install("[email protected]")
# Snapshot
renv::snapshot()
renv.lock = controlled pkg inventory.
→ renv.lock exists w/ exact vers all pkgs. renv::status() reports no issues. Every ver pinned ([email protected]), not floating.
If err: renv::install() fails specific ver → check exists CRAN archives. Use renv::install("package@version", repos = "https://packagemanager.posit.co/cran/latest") for archived.
Step 4: Version Control
git init
git add .
git commit -m "Initial validated project structure"
# Use signed commits for traceability
git config user.signingkey YOUR_GPG_KEY
git config commit.gpgsign true
→ Project under git w/ signed commits. Initial commit has validated structure + renv.lock.
If err: GPG signing fails → verify GPG key (gpg --list-secret-keys). Envs w/o GPG → document deviation + unsigned + manual audit trail in docs/change_log.md.
Step 5: IQ Protocol
validation/iq/iq_protocol.md:
# Installation Qualification Protocol
## Objective
Verify that R and required packages are correctly installed.
## Test Cases
### IQ-001: R Version Verification
- **Requirement**: R 4.5.0 installed
- **Procedure**: Execute `R.version.string`
- **Expected:** "R version 4.5.0 (date)"
- **Result**: [ PASS / FAIL ]
### IQ-002: Package Installation Verification
- **Requirement**: All packages in renv.lock installed
- **Procedure**: Execute `renv::status()`
- **Expected:** "No issues found"
- **Result**: [ PASS / FAIL ]
### IQ-003: Package Version Verification
- **Procedure**: Execute `installed.packages()[, c("Package", "Version")]`
- **Expected:** Versions match renv.lock exactly
- **Result**: [ PASS / FAIL ]
→ IQ protocol has test cases (R ver, pkg install, pkg ver verifications) w/ clear expected + pass/fail.
If err: protocol template ≠ org SOP reqs → adapt format keeping required fields (req, procedure, expected, actual, pass/fail). Consult QA for approved templates.
Step 6: Automated OQ/PQ Tests
# tests/testthat/test-analysis.R
test_that("primary analysis produces validated results", {
# Known input -> known output (double programming validation)
test_data <- read.csv(test_path("fixtures", "validation_dataset.csv"))
result <- primary_analysis(test_data)
# Compare against independently calculated expected values
expect_equal(result$estimate, 2.345, tolerance = 1e-3)
expect_equal(result$p_value, 0.012, tolerance = 1e-3)
expect_equal(result$ci_lower, 1.234, tolerance = 1e-3)
})
→ Test files exist in tests/testthat/ covering OQ (op verification each fn) + PQ (e2e validation vs independent ref). Use explicit numeric tolerances.
If err: ref vals not yet from independent calc (SAS) → placeholder w/ skip("Awaiting independent reference values") + document in traceability.
Step 7: Traceability Matrix
# Traceability Matrix
| Req ID | Requirement | Test ID | Test Description | Status |
|--------|-------------|---------|------------------|--------|
| REQ-001 | Import CSV data correctly | OQ-001 | Verify data dimensions and types | PASS |
| REQ-002 | Calculate primary endpoint | PQ-001 | Compare against reference results | PASS |
| REQ-003 | Generate report output | PQ-002 | Verify report contains all sections | PASS |
→ Matrix links every req to ≥1 test, every test to req. No orphans.
If err: untested reqs → create tests or document risk-based exclusion. Tests w/o req → link or remove out-of-scope.
Check
- Structure follows template
- renv.lock has all deps w/ exact vers
- Validation plan complete + approved
- IQ protocol executes
- OQ covers all configured fns
- PQ validates vs independent results
- Traceability matrix links reqs↔tests
- Change control process documented
Traps
install.packages()w/o pinning: Always renv w/ locked vers- No audit trail: Every change documented. Git signed commits.
- Over-validating: Risk-based. Not every CRAN pkg needs Cat 5.
- Forget system-level QA: OS + R install need IQ too
- No independent verify: PQ → compare vs results from independent (SAS, manual)
→
write-validation-documentation— detailed validation docsimplement-audit-trail— electronic records + audit trailsvalidate-statistical-output— double programming + output validationmanage-renv-dependencies— dep locking for validated envs
GitHub Repository
Related Skills
executing-plans
DesignUse the executing-plans skill when you have a complete implementation plan to execute in controlled batches with review checkpoints. It loads and critically reviews the plan, then executes tasks in small batches (default 3 tasks) while reporting progress between each batch for architect review. This ensures systematic implementation with built-in quality control checkpoints.
requesting-code-review
DesignThis skill dispatches a code-reviewer subagent to analyze code changes against requirements before proceeding. It should be used after completing tasks, implementing major features, or before merging to main. The review helps catch issues early by comparing the current implementation with the original plan.
connect-mcp-server
DesignThis skill provides a comprehensive guide for developers to connect MCP servers to Claude Code using HTTP, stdio, or SSE transports. It covers installation, configuration, authentication, and security for integrating external services like GitHub, Notion, and custom APIs. Use it when setting up MCP integrations, configuring external tools, or working with Claude's Model Context Protocol.
web-cli-teleport
DesignThis skill helps developers choose between Claude Code Web and CLI interfaces based on task analysis, then enables seamless session teleportation between these environments. It optimizes workflow by managing session state and context when switching between web, CLI, or mobile. Use it for complex projects requiring different tools at various stages.
