agentleFS
TrecekTrecek/useful-claude-skills/.claude/skills/exp-lens-reproducibility-artifacts/SKILL.md

exp-lens-reproducibility-artifacts

Skill1 starsChanged 8 months agoMITView on GitHub

Create Reproducibility Artifacts experimental design diagram showing run instructions, environment capture, data availability, determinism controls, and audit trail. Transparency lens answering "Could an independent party reproduce this?

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 6 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Focus: Run Instructions, Environment Capture, Data Availability, Determinism Controls, Audit Trail

    line 17
  2. Distinctive:

    Do not litter the codebase with useless comments, TODO markers, or explanatory annotations — the skill output and diagram speak for …

    line 30
  3. Very specific:

    BEFORE creating any diagram, LOAD the /mermaid skill using the Skill tool - this is MANDATORY

    line 37
  4. Very specific:

    Look for: requirements, Dockerfile, environment.yml, conda, pip, nix, lock

    line 49
  5. Distinctive:

    Look for: download, checksum, hash, version, dvc, data_url, manifest

    line 53
  6. Very specific:

    Look for: Makefile, run.sh, snakemake, nextflow, main, entrypoint, cli

    line 57
  7. Distinctive:

    Look for: seed, random_state, deterministic, cudnn, PYTHONHASHSEED

    line 61
  8. Distinctive:

    Look for: save, log, output, results, figures, checkpoint, wandb, mlflow

    line 65
  9. Very specific:

    Write the diagram to: temp/exp-lens-reproducibility-artifacts/exp_diag_reproducibility_artifacts_{YYYY-MM-DD_HHMMSS}.md

    line 112
  10. Distinctive:

    Before creating the diagram, verify:

    line 231

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Reproducibility Artifacts Experimental Design Lens

Philosophical Mode: Transparency Primary Question: "Could an independent party reproduce this?" Focus: Run Instructions, Environment Capture, Data Availability, Determinism Controls, Audit Trail

When to Use

  • Evaluating reproducibility of computational experiments
  • Auditing artifact completeness
  • Checking for undocumented dependencies
  • User invokes /exp-lens-reproducibility-artifacts or /make-experiment-diag reproducibility

Critical Constraints

NEVER:

  • Modify any source code files
  • Do not litter the codebase with useless comments, TODO markers, or explanatory annotations — the skill output and diagram speak for themselves

ALWAYS:

  • Trace the full chain from "clone repo" to "reproduce figures"
  • Classify every artifact as available/unavailable and versioned/floating
  • Identify the weakest link in the reproduction chain
  • Flag all silent non-determinism risks
  • BEFORE creating any diagram, LOAD the /mermaid skill using the Skill tool - this is MANDATORY

Analysis Workflow

Step 1: Launch Parallel Exploration Subagents

Spawn Explore subagents to investigate:

Environment & Dependencies

  • Find dependency files, container definitions, environment setup
  • Look for: requirements, Dockerfile, environment.yml, conda, pip, nix, lock

Data Provenance

  • Find data download scripts, checksums, versioning
  • Look for: download, checksum, hash, version, dvc, data_url, manifest

Execution Entry Points

  • Find run scripts, Makefiles, workflow managers
  • Look for: Makefile, run.sh, snakemake, nextflow, main, entrypoint, cli

Random Seed & Determinism

  • Find seed setting, nondeterminism controls
  • Look for: seed, random_state, deterministic, cudnn, PYTHONHASHSEED

Output Artifacts & Logging

  • Find result storage, logging, figure generation
  • Look for: save, log, output, results, figures, checkpoint, wandb, mlflow

Step 2: Map the Reproduction Chain

Map the full chain from "clone repo" to "reproduce figures." Identify each link:

  • Is it documented?
  • Is it automated?
  • Is it deterministic?
  • What is the weakest link?

Step 3: Classify Each Artifact Dependency

CRITICAL — Analyze Reproduction Chain: For every artifact dependency:

  • Is the source available (open vs gated)?
  • Is the version pinned or floating?
  • Is the transform deterministic?
  • Could silent environment differences change results?

Assign a status of Pass, Warn, or Fail to each link in the chain based on reproducibility confidence.

Step 4: Create the Diagram

Use flowchart with:

Direction: LR (reproduction chain flows left to right)

Subgraphs:

  • SOURCE CODE
  • ENVIRONMENT
  • DATA
  • EXECUTION
  • OUTPUTS

Node Styling:

  • cli class: Entry points and run commands
  • stateNode class: Versioned and pinned artifacts
  • handler class: Transforms and scripts
  • output class: Results and figures
  • gap class: Missing or undocumented links
  • detector class: Checksum and validation gates
  • phase class: External dependencies

Edge Labels: pinned, floating, deterministic, nondeterministic, gated

Step 5: Write Output

Write the diagram to: temp/exp-lens-reproducibility-artifacts/exp_diag_reproducibility_artifacts_{YYYY-MM-DD_HHMMSS}.md


Output Template

# Reproducibility Artifacts Diagram: {Experiment Name}

**Lens:** Reproducibility Artifacts (Transparency)
**Question:** Could an independent party reproduce this?
**Date:** {YYYY-MM-DD}
**Scope:** {What was analyzed}

## Artifact Inventory

| Artifact | Available? | Versioned? | Deterministic? |
|----------|------------|------------|----------------|
| {artifact} | {Yes/No/Gated} | {Pinned/Floating/None} | {Yes/No/Unknown} |

## Reproduction Chain Diagram

```mermaid
%%{init: {'flowchart': {'nodeSpacing': 50, 'rankSpacing': 60, 'curve': 'basis'}}}%%
flowchart LR
    %% CLASS DEFINITIONS %%
    classDef cli fill:#1a237e,stroke:#7986cb,stroke-width:2px,color:#fff;
    classDef stateNode fill:#004d40,stroke:#4db6ac,stroke-width:2px,color:#fff;
    classDef handler fill:#e65100,stroke:#ffb74d,stroke-width:2px,color:#fff;
    classDef phase fill:#6a1b9a,stroke:#ba68c8,stroke-width:2px,color:#fff;
    classDef newComponent fill:#2e7d32,stroke:#81c784,stroke-width:2px,color:#fff;
    classDef output fill:#00695c,stroke:#4db6ac,stroke-width:2px,color:#fff;
    classDef detector fill:#b71c1c,stroke:#ef5350,stroke-width:2px,color:#fff;
    classDef gap fill:#ff6f00,stroke:#ffa726,stroke-width:2px,color:#000;
    classDef integration fill:#c62828,stroke:#ef9a9a,stroke-width:2px,color:#fff;

    subgraph Source ["SOURCE CODE"]
        REPO["Git Repository<br/>━━━━━━━━━━<br/>Commit hash<br/>Branch"]
        ENTRY["Entry Point<br/>━━━━━━━━━━<br/>run.sh / Makefile"]
    end

    subgraph Env ["ENVIRONMENT"]
        DEPS["Dependencies<br/>━━━━━━━━━━<br/>requirements.txt / lock"]
        MISSING["Undocumented Dep<br/>━━━━━━━━━━<br/>No version pin"]
        EXTDEP["External Service<br/>━━━━━━━━━━<br/>API / cloud resource"]
    end

    subgraph Data ["DATA"]
        RAW["Raw Dataset<br/>━━━━━━━━━━<br/>Checksum available?"]
        CHKSUM["Checksum Gate<br/>━━━━━━━━━━<br/>sha256 / md5"]
        GATED["Gated Dataset<br/>━━━━━━━━━━<br/>Access required"]
    end

    subgraph Exec ["EXECUTION"]
        SEED["Seed Control<br/>━━━━━━━━━━<br/>PYTHONHASHSEED<br/>random_state"]
        SCRIPT["Pipeline Script<br/>━━━━━━━━━━<br/>Deterministic?"]
    end

    subgraph Outputs ["OUTPUTS"]
        RESULTS["Results / Metrics<br/>━━━━━━━━━━<br/>Logged?"]
        FIGS["Figures<br/>━━━━━━━━━━<br/>Reproducible?"]
    end

    %% REPRODUCTION CHAIN %%
    REPO -->|"pinned"| ENTRY
    ENTRY -->|"loads"| DEPS
    DEPS -.->|"floating"| MISSING
    MISSING -.->|"nondeterministic"| SCRIPT
    EXTDEP -->|"gated"| SCRIPT
    RAW -->|"verify"| CHKSUM
    CHKSUM -->|"deterministic"| SCRIPT
    GATED -.->|"gated"| SCRIPT
    SEED -->|"controls"| SCRIPT
    SCRIPT -->|"produces"| RESULTS
    SCRIPT -->|"generates"| FIGS

    %% CLASS ASSIGNMENTS %%
    class REPO,ENTRY cli;
    class DEPS,RAW stateNode;
    class SCRIPT,SEED handler;
    class EXTDEP phase;
    class RESULTS,FIGS output;
    class CHKSUM detector;
    class MISSING,GATED gap;

Color Legend:

ColorCategoryDescription
Dark BlueEntry PointRun commands and source code
TealVersioned ArtifactPinned dependencies and checksummed data
OrangeTransform / ScriptPipeline scripts and execution steps
PurpleExternal DependencyExternal services and APIs
Dark TealOutputResults, metrics, and figures
RedValidation GateChecksum and integrity checks
AmberMissing LinkUndocumented or gated dependencies

Reproduction Checklist

Step-by-step instructions with pass/fail status:

  • Clone repository at pinned commit
  • Reproduce environment from lock/container file
  • Download data and verify checksums
  • Set all random seeds as documented
  • Execute pipeline via documented entry point
  • Compare output metrics/figures to reported values

Weakest Links

LinkIssueSeverityRecommendation
{link}{undocumented/gated/floating/nondeterministic}{High/Medium/Low}{action}

---

## Pre-Diagram Checklist

Before creating the diagram, verify:

- [ ] LOADED `/mermaid` skill using the Skill tool
- [ ] Using ONLY classDef styles from the mermaid skill (no invented colors)
- [ ] Diagram will include a color legend table

---

## Related Skills

- `/make-experiment-diag` - Parent skill for lens selection
- `/mermaid` - MUST BE LOADED before creating diagram
- `/exp-lens-pipeline-integrity` - For data leakage audit
- `/exp-lens-variance-stability` - For result stability across seeds

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.