Skip to main content
prompt GP ChatGPT official · Anthropic

sim_critic — Claude Code prompt: model audit

The sim_critic Claude Code prompt audits a Monte Carlo clinical-trial model against its evidence, code, results, and report. It flags calibration issues and returns a verdict with ranked change requests.

no license ★ 654 updated 2026-11-09

Install into Claude Code · Plugin marketplace All methods and platforms ↓
/plugin marketplace add anthropics/life-sciences
/plugin install clinical-trial-forecast@life-sciences
Open on GitHub

What sim_critic does

The prompt reviews a Monte Carlo model for a clinical trial. It checks parameter estimates against source evidence, examines statistical logic and simulation results, and assesses reproducibility and report clarity. It also looks for duplicate adjustments and shifts in either direction.

The review uses project files such as sim/params.json, sim/workspace/simulate.py, sim/workspace/results.json, plus evidence from intake/, knowledge/, topics/, analyses/, and research/. It works by reading files only; it does not access the web or run code.

The output contains section scores, ranked requests for changes, open research questions, and a critique in Markdown of up to 1,500 words. Its verdict is accept, revise_params, revise_code, or revise_both.

What it helps with

  • compare model parameters with trial evidence
  • detect overlapping adjustments that distort PTS
  • check simulation logic and reproducibility
  • rank changes needed in the model or report

Example prompts for the agent

  • Audit the clinical-trial model using the project files and return a verdict with ranked requests.
  • Check whether any parameters count the same effect twice and assess the impact on PTS.
  • Compare the simulation with the evidence and write a critique for the committee.

How to install sim_critic

1Plugin marketplace

/plugin marketplace add anthropics/life-sciences
/plugin install clinical-trial-forecast@life-sciences

2Manually: put the files in place

.claude/commands/sim_critic.md

Questions about sim_critic

How do I install sim_critic in Claude Code?

Add the marketplace with /plugin marketplace add anthropics/life-sciences, then install the plugin with /plugin install clinical-trial-forecast@life-sciences. The command file is at .claude/commands/sim_critic.md.

What does sim_critic do?

It audits a clinical-trial Monte Carlo model, covering its parameters, code, results, and reproducibility. The response includes a verdict, ranked change requests, and research questions.

Which files does sim_critic review?

It reads model files under sim/ and supporting evidence from intake/, knowledge/, topics/, analyses/, and research/. It does not execute code or access the web.

Similar prompts

All prompts for Science and research →

From the same set29

The whole set →
trial_checklist_worker The trial_checklist_worker Claude Code prompt analyzes one clinical trial against a fixed question group. It draws on registry records, project files, and research sources. GP ChatGPT ★ 654repo stars trial_checklist_condense The trial_checklist_condense Claude Code prompt condenses one clinical trial’s item-level analysis into a report and a searchable card. It draws from analysis files and a trial registry summary. GP ChatGPT ★ 654repo stars sim_spec The sim_spec Claude Code prompt turns assembled trial evidence into a parameter sheet for a Monte Carlo model. A separate builder implements that sheet. GP ChatGPT ★ 654repo stars sim_finalize The sim_finalize prompt for Claude Code prepares a report on an accepted clinical-trial simulation. Run it to assemble the model description, summary measures, and figures. GP ChatGPT ★ 654repo stars sim_build The sim_build Claude Code prompt builds and documents a Monte Carlo simulation of clinical trial outcomes. It models program success while accounting for parameter uncertainty and operational failure modes. GP ChatGPT ★ 654repo stars sim_brief The sim_brief Claude Code prompt guides a clinical-trial simulation and its report. It runs when an agent is asked to implement a model from params.json. GP ChatGPT ★ 654repo stars researcher researcher is a Claude Code prompt for answering one research question about an indication. It draws on project files and external sources, checks figures, and returns cited findings. GP ChatGPT ★ 654repo stars research_planner The research_planner Claude Code prompt plans the next research round for a report on a medical indication. It prioritizes questions that could shift a recommendation or narrow decision-critical estimates. GP ChatGPT ★ 654repo stars

Project sponsors

Companies that keep this analytics open