Pharma Stats Skill — Development skill for Claude Code
Claude Code skill and project template for pharma biostatistics: human-in-the-loop SDTM to ADaM to TLF workflows in SAS/R, with re-read-before-edit, user-runs-code, decision logs, and QC commands and.
How to install Pharma Stats Skill
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open wal615/pharma-stats-skill and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Pharma Stats Skill does
Claude Code skill and project template for pharma biostatistics: human-in-the-loop SDTM to ADaM to TLF workflows in SAS/R, with re-read-before-edit, user-runs-code, decision logs, and QC commands and agents.
Alternatives in Development
- Obsidian MCP Server — Read, write, search, and surgically edit Obsidian vault notes, tags, and frontmatter via MCP 686 ★
- MCP Edit — tilth_write replaces the host Edit and Write tools 349 ★
- Cloud Run Debug — Diagnose a failing Cloud Run service — Antigravity (agy/Gemini) digests the error logs cheaply, Claude infers 284 ★
README
Pharma Stats Skill for Claude Code
A Claude Code skill and project template for biostatisticians doing clinical trial statistical programming (SDTM → ADaM → TLF) in SAS or R.
The idea
End-to-end AI derivation fails because a derivation is a chain of small hidden assumptions, and the agent guesses each one silently. This skill flips the order: **decide first, derive second.** The statistician makes the decisions and runs the code; Claude Code writes small, self-contained snippets. Every decision is visible in the data before it becomes code, and recorded in a decision log.
What's inside
SKILL.md— the skill itself: the hybrid loop, the non-negotiable rules, and how to use the commands and agents.templates/— a drop-in project template:- Layered
CLAUDE.mdfiles (root +adam/+tlf/+sdtm/+specs/) docs/STUDY_MAP.md— project index, maintained by handdocs/DECISIONS.md— decision log / audit trail.claude/commands/—/check-log,/qc-tlf,/new-adam.claude/agents/—study-explorer,tlf-reviewer.claude/settings.json— permissions (Rscript/SAS runs need approval) and hooks (a reminder to re-read a file before every edit)
- Layered
Use it as a skill
Copy this folder to `~/.claude/skills/pharma-stats/` (all projects) or to a project's `.claude/skills/pharma-stats/`, then invoke it with `/pharma-stats` or let Claude Code pick it up when the task matches.
Use it as a project template
Copy the contents of `templates/` into your study project root and replace every `[placeholder]` with the real study information. Start with `study-explorer` to map a new study into `docs/STUDY_MAP.md`.
Core rules (enforced by the template)
- Re-read a file before editing it — the user edits by hand in RStudio, and a remembered version may be stale.
- The user runs the code by default; Claude Code runs it only when explicitly asked.
- Every snippet is self-contained: it loads its own libraries and da
Related Skills
Router Stats
Show skill-router match statistics from ~/.claude/logs/intent-router-.jsonl and prune logs older than 30 days.
Clear Rewrite
Rewrite text to pass CLEAR-100. Plain words, human voice, real structure, no filler. Runs the checker in a loo
Add Human Approval
Scaffold a human-in-the-loop approval gate into an agent so it pauses before a consequential action and resume
The Docket
the-docket is a Claude Code plugin that keeps a repo's decisions from being undone. A hook prints what governs
AgentsUsage
A lightweight Windows system tray application that parses local Claude CLI and Codex usage logs, hosting an em
Classify Stats
Render classification statistics (1/7/30/90 days, human vs agent pyramid + table)
Related Agents
Operador
Operador de infraestructura local del Framework Bot en CachyOS + Omarchy. Maneja Docker/compose (stop, up -d -
AI Clinical Reviewer
Clinical AI / clinical decision-support pre-implementation reviewer. Specialises in FDA GMLP (10 guiding princ
Swarm Member
Read-only swarm team member. Spawned by swarm via the Agent tool. Tools restricted to read and research only —