From claudecode-research-harness-workflow
Initializes a research project with folder structure, study_spec.md, and analysis_plan.md. Runs once at project start; does not analyze or audit data.
How this skill is triggered — by the user, by Claude, or both
Slash command
/claudecode-research-harness-workflow:research-harness-setup [--study-name NAME] [--data-path PATH][--study-name NAME] [--data-path PATH]This skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Initialize a new empirical research project with source-of-truth files, a controlled folder structure, and data protection rules.
Initialize a new empirical research project with source-of-truth files, a controlled folder structure, and data protection rules.
This skill runs once at the start of a project. It does not run analysis, audit data, or claim feasibility.
| Input | Action |
|---|---|
/research-harness-setup | Interactive: prompt for study name and data path, then create all files |
/research-harness-setup --study-name "X" --data-path "data/raw/X.csv" | Non-interactive: use provided values |
If arguments are not provided, ask the user for:
unknown if data not yet available)unknown)Do not invent answers. If the user says unknown, write unknown in the spec.
Create the following directories if they do not exist. Do not delete or overwrite existing content.
data/raw/
data/intermediate/
data/processed/
scripts/
analysis/
output/
logs/
reports/
Copy templates/study_spec.md to study_spec.md at the project root.
Fill in:
Leave all other fields as unknown. The researcher fills in the full spec before Stage 2.
If study_spec.md already exists: read it, confirm with the user before overwriting, and preserve existing content unless the user explicitly requests replacement.
Copy templates/analysis_plan.md to analysis_plan.md at the project root.
Fill in:
study_spec.mdLeave all task fields as cc:todo with unknown paths.
If analysis_plan.md already exists: do not overwrite. Notify the user.
Append the following block to the project .gitignore if it does not already contain it:
# Research Agent Harness — raw data is read-only
# Never commit cleaned or processed datasets unless explicitly approved
data/processed/
data/intermediate/
logs/
Write a data/raw/READONLY.md file with the following content:
# data/raw — READ-ONLY
This folder contains raw, unmodified data files.
No script, agent, or tool may write, edit, rename, or delete files in this folder.
All data transformations must produce new files in data/processed/ or data/intermediate/.
See docs/INTEGRITY-RULES.md Rule 1.
Print a setup summary:
Research Agent Harness — Setup Complete
Study spec: study_spec.md [created / already existed]
Analysis plan: analysis_plan.md [created / already existed]
Folders: data/raw, data/intermediate, data/processed,
scripts, analysis, output, logs, reports
Data protection: data/raw/READONLY.md written
Next step: Fill in study_spec.md, then run /research-harness-audit
data/raw/study_spec.md without user confirmation if it already existsstudy_spec.md exists with study name, data path, and research question filled in (or unknown)analysis_plan.md exists with header fields populateddata/raw/READONLY.md existsdata/raw/ were modifiedTell the user:
Fill in the remaining fields in
study_spec.md— particularly identification strategy, variables, and sample restrictions — before running/research-harness-audit. Whenstudy_spec.mdis complete and you have approved it, run/research-harness-audit.
npx claudepluginhub maxwell2732/claudecode-research-harness-workflow --plugin claudecode-research-harness-workflowGenerates an executable empirical analysis plan from study_spec.md, audit report, and cleaned data structure. Outputs analysis_plan.md for human approval before analysis execution.
Installs OMCR project infrastructure: CLAUDE.md markers, agent memory directories, bibliography stubs, and a permissions allowlist. Safe to re-run.
7-step setup wizard that produces a complete, ready-to-run research.md without executing the research loop. Walks the user through goal, metric, search space, constraints, evaluator design, and baseline measurement, then writes the file. TRIGGER when: user wants to set up a research project; user wants to plan before running the loop; user says "plan my research"; user has a goal but no research.md; user invokes /autoresearch:plan. DO NOT TRIGGER when: research.md already exists and the user wants to run the loop; user wants a one-shot answer; user wants to debug, not optimize.