agentleFS
arthur-plautzarthur-plautz/research-toolkit/.claude/skills/analysis/SKILL.md

analysis

Skill0 starsChanged 4 months agoMITView on GitHub

Run analysis scripts for the exploration/analysis pipeline. Guides the user through selecting one or more analyses, collecting experiment/seed/path arguments, and executing the corresponding scripts. Reuses existing CSV data by default; only reruns scripts when explicitly asked.

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 5 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If the user has not already specified which analysis they want, show the catalog and ask them to choose.

    line 20
  2. Distinctive:

    Always: do this step, even if the user gave a hint.

    line 33
  3. Very specific:

    Run whatever lookup finds the project's experiment output directories (e.g.

    line 37
  4. Distinctive:

    After experiments are selected, check which seeds/runs exist for the first selected experiment and ask the user to confirm which to use …

    line 52
  5. Very specific:

    Before running any script, check if a CSV for this analysis already exists in analysis/figures/.

    line 58
  6. Distinctive:

    Always pass the confirmed experiment/seed selection explicitly — never rely on script defaults for these.

    line 70
  7. Distinctive:

    Do not link to or mention any generated image

    line 83

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

Purpose

This skill runs analysis scripts from analysis/scripts/. It always asks the user to confirm experiments/runs and seeds before running anything, reuses cached CSV data when it exists, and displays tables inline rather than generating images.

For background on experiment types, data output formats, and metric definitions, see docs/project/analysis.md.

This skill's catalog is a placeholder. Populate the table in Step 1 with this project's actual analysis scripts (name, what it produces, single- vs. multi-experiment) as they're added under analysis/scripts/.


Workflow

Step 1 — Present the catalog

If the user has not already specified which analysis they want, show the catalog and ask them to choose. Accept names, numbers, or natural descriptions.

[1]  {script_name}   {what it produces}   (single experiment | per experiment)
[2]  ...

The user may select multiple analyses.


Step 2 — Always ask for experiments and seeds

Always do this step, even if the user gave a hint. Do not assume or default silently.

2a — Discover available experiments/runs

Run whatever lookup finds the project's experiment output directories (e.g. find data/ -name "*.npy" | ...) and present them as a numbered list:

Available experiments:
  [1] {experiment-0}
  [2] {experiment-1}
  ...

Which experiments do you want to include? (enter numbers or names)

For single-experiment-only analyses, ask them to pick one.

2b — Confirm seeds/runs

After experiments are selected, check which seeds/runs exist for the first selected experiment and ask the user to confirm which to use (all, or a subset).


Step 3 — Check for existing data

Before running any script, check if a CSV for this analysis already exists in analysis/figures/. If a matching CSV exists, reuse it by default — skip script execution and go directly to Step 4. Only regenerate if the user explicitly asked to rerun or regenerate the analysis.


Step 4 — Run the script (if needed)

Scripts must be run from analysis/scripts/, e.g.:

cd analysis/scripts && source ../../venv/bin/activate && python <script>.py <args>

Always pass the confirmed experiment/seed selection explicitly — never rely on script defaults for these.


Step 5 — Display results

For table analyses:

  • Display inline using the skill's helper script:
source venv/bin/activate && python .claude/skills/analysis/scripts/display_table.py \
  analysis/figures/<csv_file>.csv [analysis/figures/<csv_file2>.csv ...]
  • Pass all relevant CSVs in one call to print them sequentially
  • Do not link to or mention any generated image

For visual analyses:

  • Confirm the saved image path
  • Report: Saved → analysis/figures/<filename>.png

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.