ds-review
This skill should be used when running Phase 4 of the /ds workflow to review methodology, data quality, and statistical validity. Provides structured review checklists, confidence scoring, and issue identification for data analysis validation.
npx skills add majiayu000/claude-skill-registry --skill ds-review-edwinhu-workflows-2 --agent claude-code
Same command for any agent — swap --agent for codex, cursor, copilot.
Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.
What it does
The skill provides a structured review framework for Phase 4 of a /ds workflow, focusing on methodology, data quality, and statistical validity. It offers: structured review checklists, a confidence scoring scheme, and identification of issues related to data analysis validation. It supports selectable review strategies (Single reviewer by default or Parallel review with three specialized reviewers) and includes procedures for spawning reviewers, prerequisites checks, and rigorous reporting constraints.
How it works
- Announce the use of ds-review in Phase 4.
- Present a Review Strategy Choice to decide between:
- Single reviewer (Default): continue with a consolidated review following The Iron Law of DS Review.
- Parallel review (Research-grade): enable CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS and spawn three reviewers (Methodology, Reproducibility, Code Quality) after preparing inline context variables (ANALYSIS_FILES, SPEC_CONTEXT, PLAN_TASKS, LEARNINGS_PIPELINE, PLUGIN_ROOT).
- If Parallel review is chosen, perform a Prerequisites Check ensuring SPEC.md, PLAN.md, LEARNINGS.md exist and Analysis files are identified; otherwise stop.
- In the Methodology Review section, focus on statistical methodology, assumptions, sample size, multiple testing, causality, bias, confounders, Simpson’s paradox, and related checks.
- Enforce the Iron Law of DS Review to report issues only with confidence >= 80, with a strict process for validating conclusions before reporting.
- Provide a detailed output structure for reporting: Critical Methodology Issues, Important Methodology Issues, and a Methodology Summary with verdict (APPROVED or CHANGES REQUIRED).
- Include extensive guidance blocks for Red Flags, Confidence Scoring, and Rationalization Prevention to avoid non-evidenced conclusions.
- Include a Decorative but strict 'Required Output Structure' that outlines how to format the final markdown report.
When to use it
Use this skill for Phase 4 reviews when methodology validity, data quality, and statistical soundness are critical, especially for publication-bound analyses, high-stakes decisions, or regulatory/compliance contexts. Use Single reviewer unless the user selects Parallel review (Research-grade).
What it can touch
- Prerequisites and review workflow orchestration. It references and uses prompts and contextual blocks (ANALYSIS_FILES, SPEC_CONTEXT, PLAN_TASKS, LEARNINGS_PIPELINE) to initialize reviewers.
- It requires CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS for parallel review; otherwise it falls back to a single reviewer.
Caveats
- Requires explicit prerequisites to be present; otherwise returns to /ds-implement.
- The Iron Law of DS Review enforces reporting only issues with confidence >= 80; lower-confidence observations are not reported.
- Parallel review relies on spawning three specialized reviewers and assumes ability to substitute inline context blocks.
Announce: "Using ds-review (Phase 4) to check methodology and quality." ## Review Strategy Choice After announcing phase, choose review strategy. **Skip this choice when:** - Exploratory analysis (one-off, not for publication) - Trivial changes (formatting, documentation) - Internal reporting (low-stakes, quick turnaround) - Single notebook with < 100 LOC **Otherwise, ask the user:** ```python AskUserQuestion(questions=[{ "question": "How should we review this analysis?", "header": "Review Strategy", "options": [ {"label": "Single reviewer (Default)", "description": "Combined review covering methodology, data quality, and reproducibility. Faster, lower overhead."}, {"label": "Parallel review (Research-grade)", "description": "Spawn 3 specialized reviewers (Methodology, Reproducibility, Code quality). Use for publications, high-stakes decisions, or research-grade work. Requires reconciliation."} ], "multiSelect": false }]) ``` **If Single reviewer:** Proceed to [The Iron Law of DS Review](#the-iron-law-of-ds-review) below (current behavior). **If Parallel review:** Skip to [Parallel Review (Research-Grade)](#parallel-review-research-grade). --- ## Parallel Review (Research-Grade) Us
- Review Strategy Choice
- Parallel Review (Research-Grade)
- 1. Prerequisites Check
- 2. When to Use Parallel Review
- 3. Create Team and Spawn Reviewers
- Methodology Review
- Critical Methodology Issues (Confidence >= 90)
- Important Methodology Issues (Confidence 80-89)
- Methodology Summary
- Reproducibility Review
- Reproduction Attempt
- Critical Reproducibility Issues (Confidence >= 90)
- Important Reproducibility Issues (Confidence 80-89)
- Reproducibility Summary
What does the ds-review skill do?
This skill should be used when running Phase 4 of the /ds workflow to review methodology, data quality, and statistical validity. Provides structured review checklists, confidence scoring, and issue identification for data analysis validation.
How do I install it?
Run `npx skills add majiayu000/claude-skill-registry --skill ds-review-edwinhu-workflows-2 --agent claude-code` — it drops the skill into your project so the agent can pick it up. Swap the --agent value for codex, cursor or copilot if you use one of those.
Where does this skill come from?
From majiayu000/claude-skill-registry, a repository with 534 stars. We read it straight from the repository tree rather than a submitted listing, so what you see here is what is actually published.
Is a popular skill a good skill?
Not necessarily. Stars measure attention, not adoption — a repository can trend for a week and be abandoned. That is why we show the weekly change from our own snapshots next to the total, instead of a single flattering number.