agentleFS
robium-airobium-ai/robium/skills/test-assets/SKILL.md

test-assets

Skill93 starsChanged 3 months agoMITView on GitHub

Choose and manage reproducible worlds, models, datasets, and recordings for robotics tests.

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:

    Start from the behavior under test, then select the smallest representative world, model, recording, dataset slice, checkpoint, or golden.

    line 13
  2. Distinctive:

    Prefer a recognizable public asset when it matches the behavior.

    line 15
  3. Distinctive:

    Add an asset only when an active test consumes it.

    line 17
  4. Distinctive:

    Use a pinned pointer for large or independently maintained assets.

    line 24
  5. Very specific:

    Use examples/catalog.yaml only as a proven catalog shape, not as a universal asset set.

    line 41
  6. Very specific:

    Use scripts/fetch_assets.py only for checksum-pinned pointer archives that match its supported manifest shape.

    line 43
  7. Distinctive:

    Every fixture has an active consumer, reproducible source or generator, license evidence, and an assertion whose tolerance matches how the …

    line 50

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

Test assets

A fixture earns its place by being a pinned input with a known consumer, not by being realistic or popular.

Choose what the test needs

  • Start from the behavior under test, then select the smallest representative world, model, recording, dataset slice, checkpoint, or golden.
  • Prefer a recognizable public asset when it matches the behavior. It gives reviewers context and avoids maintaining an equivalent private fixture.
  • Add an asset only when an active test consumes it. Record the upstream revision, license, verification method, entrypoint, and derivation.
  • Reuse outputs from verified scenarios when useful: a generated map or bag can become the pinned input to the next layer.

Choose how it is owned

  • Use a pinned pointer for large or independently maintained assets. Verify an immutable revision and checksum, and cache the fetched bytes.
  • Vendor only when the test must survive upstream availability or inspect the exact bytes, redistribution is allowed, and the project has a stated size budget.
  • Generate project-specific bags and goldens from a committed, seeded producer when no canonical source exists.
  • Compare re-simulated behavior with tolerance bands. Reserve byte equality for pure replay or deterministic transforms.

Go deeper only when needed

  • To select a known world, robot model, dataset, or recording, read references/canonical-assets.md and verify its dated evidence against the current source before adoption.
  • To define a project catalog, manifest, slice, or golden, read references/test-assets-layout.md.
  • Use examples/catalog.yaml only as a proven catalog shape, not as a universal asset set.
  • Use scripts/fetch_assets.py only for checksum-pinned pointer archives that match its supported manifest shape.
  • Testing owns the test layers and pass bars. Data owns training-data sourcing; the simulator and Hub skills own loading and transfer mechanics.

Done

  • Every fixture has an active consumer, reproducible source or generator, license evidence, and an assertion whose tolerance matches how the fixture is produced.

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.