Help us improve
Share bugs, ideas, or general feedback.
From claudecode-research-harness-workflow
Generates 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.
npx claudepluginhub maxwell2732/claudecode-research-harness-workflow --plugin claudecode-research-harness-workflowHow this skill is triggered — by the user, by Claude, or both
Slash command
/claudecode-research-harness-workflow:research-harness-plan [--update] [--add STAGE][--update] [--add STAGE]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
Generate an executable empirical analysis plan based on the approved study specification, audit findings, and cleaned data structure.
Creates concrete analysis plans from approved research questions, covering model specs, confounds, power, and pipeline structure. Use before touching outcome data or fitting models.
Executes approved analysis tasks from analysis_plan.md by writing reproducible scripts, running them, saving logs and outputs, and updating task statuses.
Phase 2 of the /ds workflow — profiles data and creates a task breakdown from a spec. Requires SPEC.md from ds-brainstorm.
Share bugs, ideas, or general feedback.
Generate an executable empirical analysis plan based on the approved study specification, audit findings, and cleaned data structure.
This skill runs after /research-harness-clean and before /research-harness-work. The output — analysis_plan.md — requires human approval before any analysis script is run.
| Input | Action |
|---|---|
/research-harness-plan | Generate a fresh analysis_plan.md from study_spec.md + reports |
/research-harness-plan --update | Update an existing analysis_plan.md with new tasks or revised scope |
/research-harness-plan --add robustness | Add a robustness-check stage to an existing plan |
Before writing any plan content:
study_spec.md. If it does not exist or is substantially empty, stop and tell the user to run /research-harness-setup first.reports/data_audit_report.md. If it does not exist, stop and tell the user to run /research-harness-audit first.reports/data_cleaning_report.md. If it does not exist, stop and tell the user to run /research-harness-clean first.If any check fails: report which check failed and stop. Do not generate a plan for data that does not exist.
Read and synthesize:
study_spec.md: research question, identification strategy, outcome, treatment, covariates, sample restrictions, expected outputsreports/data_audit_report.md: available variables, sample sizes, ID structure, missingness, feasibility assessmentreports/data_cleaning_report.md: final row count, variable names in the analysis-ready dataset, any unresolved issuesIdentify:
unknown in the planfeasible or partially feasibleDo not propose analysis tasks that depend on variables not found in the cleaned data. If a key variable is missing, note the task as cc:blocked with the reason.
The plan must follow this stage structure:
study_spec.md §2study_spec.md §6For each task, specify:
1.1, 2.1, 3.1)analysis/<stage>/scripts/<task>.R)analysis/<stage>/logs/<task>.log)output/tables/<task>.csv or output/figures/<task>.pdf)cc:todoFor every task in the main models and robustness stage:
study_spec.md §2 into the task description[descriptive], [correlational], [quasi-experimental: ...], or [experimental][experimental]study_spec.mdAt the end of the plan, include a section:
Limitations:
Unresolved questions:
Do not invent answers to unresolved questions. Leave them as unknown and mark the dependent tasks cc:blocked.
Write the complete analysis_plan.md to the project root using templates/analysis_plan.md as the structure.
If analysis_plan.md already exists and --update was not passed: ask the user whether to overwrite or append. Do not silently overwrite an existing plan.
Print a plan summary:
Research Harness Plan — Draft Complete
Tasks generated:
Stage 1 (Descriptive): N tasks
Stage 2 (Main models): N tasks
Stage 3 (Robustness): N tasks
Stage 4 (Heterogeneity): N tasks
Stage 5 (Figures/tables): N tasks
Blocked (missing data): N tasks
Limitations: [count]
Unresolved questions: [count]
⚠ HUMAN APPROVAL REQUIRED before running /research-harness-work.
Review analysis_plan.md and confirm the plan is correct before proceeding.
study_spec.mdcc:done — all tasks start as cc:todo or cc:blockedanalysis_plan.mdanalysis_plan.md exists at the project rootanalysis_plan.md written with all stagesTell the user:
Review
analysis_plan.md. When you are satisfied with the plan, approve it.Then run
/research-harness-work 1.1(or the task number you want to execute first).Do not run
/research-harness-workuntil you have reviewed and approved the plan.