maxwell2732Community·32 Agent Skills Included

claudecode-research-harness-workflow

Auditable, reproducible empirical research workflows from raw data to replication package

Guides empirical research through seven controlled stages: setup, data audit, cleaning, planning, analysis, review, and release. Prevents fabricated results by requiring a script, log, and output file for every reported number, and keeps raw data read-only. Produces merge reports, review verdicts, and complete replication packages so every table and figure is traceable to its source.

39|46

↗

npx skills add maxwell2732/claudecode-research-harness-workflow --all -g -y

Available:

Gives the agent strict operating rules for empirical research: never modify raw data, never invent numbers, log every script run, document every merge, and stop instead of fabricating results when data is insufficient.

All Skills in This Repository (32)

Pure Emerald Level Indicators

📦 In Repo

maxwell2732

session-control

Apply /work --resume and --fork flags to update session state files.

ui

Generate accessible UI components like hero sections and contact forms.

session-state

Validate and apply session state transitions across workflow phase boundaries.

agent-browser

Automate browser navigation, form filling, and element verification via the agent-browser CLI.

workflow-guide

Formalize role assignments and handoff procedures for Cursor PM and Claude Code workflows.

research-harness-work

Execute approved analysis tasks by generating reproducible scripts and maintaining logs.

harness-loop

Coordinate long-running task loops with fresh context and controlled termination.

memory

Manages and queries SSOT memory and cross-tool recall for decisions, patterns, learnings.

harness-sync

Reconcile Plans.md with commits to surface implementation drift.

auth

Implement authentication and payments with Clerk, Supabase Auth, or Stripe.

ci

Diagnose CI/CD failures and output structured fix recommendations.

deploy

Automate deployment setup and health monitoring for Vercel or Netlify projects.

Community

Intermediate

Frequently Asked Questions

FAQPage Schema

How to install claudecode-research-harness-workflow?▼

Run `npx skills add maxwell2732/claudecode-research-harness-workflow --all -g -y` in your terminal to install all skills globally.

What problem does the research harness solve?▼

It stops agents from producing unverifiable research results by enforcing the rule: no script, no log, no claim. Every number in your paper must trace back to a runnable script and its log file.

How do I clean and merge survey data with this workflow?▼

Place raw files in data/raw/, then run /research-harness-audit followed by /research-harness-clean. The agent writes reproducible cleaning scripts, logs every dropped row, and documents every merge with pre and post row counts.

Can it replicate a published paper's sample and variables?▼

Yes. A dedicated mode extracts an analysis-ready subsample from raw survey panels matching a published paper's sample filters and variable definitions, and generates a codebook comparing your means against the paper's reported statistics.

Do I need coding experience to use this research workflow?▼

No. You describe your study in plain language and approve each stage; the agent writes and runs the R, Stata, or Python scripts while enforcing reproducibility rules automatically.