robium / skills
AGENTS.md93 starsChanged 3 months agoMITView on GitHub
This file applies under skills/. Load skill-author before changing skill content and learning-loop when working from captured evidence. - Only architect needs the full catalog. Other skills name direct neighbors only where work crosses a real boundary. - Live frontmatter contains only name and a short description. - Give each entrypoint one strong organizing idea and keep it human-readable. There is no required section template. - Assume ordinary coding competence. Include only guidance that changes a capable agent's decisions. -…
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,774 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository32 other files it gives its agents.
- AGENTS.mdAGENTS.md
- learnings/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- learnings/CLAUDE.mdCLAUDE.md
- skills/CLAUDE.mdCLAUDE.md
- app-publishingSkill
- architectSkill
- cloud-runSkill
- dataSkill
- environmentsSkill
- foxgloveSkill
- gazeboSkill
- gemini-roboticsSkill
- huggingfaceSkill
- integrationSkill
- isaac-labSkill
- isaac-simSkill
- learning-loopSkill
- lerobotSkill
- live-demoSkill
- miningSkill
- mujocoSkill
- navigationSkill
- rerunSkill
- ros2Skill
- runpodSkill
- rviz2Skill
- simulationSkill
- skill-authorSkill
- test-assetsSkill
- testingSkill
- visualizationSkill
The file
We split this AGENTS.md into the rules it gives an agent
5 rules. About 0 are common, 3 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
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
- Very specific: line 8
Only architect needs the full catalog.
- Distinctive: line 15
Put conditional commands, failures, tuning, platform evidence, schemas, and substantial examples in focused support files.
- Distinctive: line 18
Keep official upstream documentation authoritative for volatile commands, APIs, packages, and configuration.
- Very specific: line 29
Run uv run skills/skill-author/scripts/validate_skills.py once after each coherent batch of live skill changes, not after every file or …
- Distinctive: line 31
Use trigger or task evals only when they protect a meaningful routing or behavior regression.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md 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.
Files
Skill guidance
This file applies under skills/. Load skill-author before changing skill
content and learning-loop when working from captured evidence.
Keep discovery and context cheap
- Only
architectneeds the full catalog. Other skills name direct neighbors only where work crosses a real boundary. - Live frontmatter contains only
nameand a shortdescription. - Give each entrypoint one strong organizing idea and keep it human-readable. There is no required section template.
- Assume ordinary coding competence. Include only guidance that changes a capable agent's decisions.
- Put conditional commands, failures, tuning, platform evidence, schemas, and substantial examples in focused support files. Link them where they become relevant instead of loading them by default.
- Keep official upstream documentation authoritative for volatile commands, APIs, packages, and configuration.
- Preserve exact Robium-observed values with their conditions; never present one application's workaround as a universal default.
Change and review
- Git history is the normal record of live skill changes. Do not add per-skill versions, changelogs, README files, or archive snapshots.
- Preserve useful examples and scripts only when they provide reusable value; validate changed executable artifacts proportionally.
- Run
uv run skills/skill-author/scripts/validate_skills.pyonce after each coherent batch of live skill changes, not after every file or sentence. - Use trigger or task evals only when they protect a meaningful routing or behavior regression. Do not test headings or exact prose.
- Regenerate
cli/src/catalog.jsonafter a skill name or description changes.
Robium ships practical robotics knowledge and real reusable examples, not an application framework or invented DSL.
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
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.