kumo-engine / development
AGENTS.md34 starsChanged 23 days agoApache-2.0View on GitHub
You can start at the repository root or in a component directory. Read the applicable AGENTS.md files and follow their links when a change spans components. Component instructions add local constraints; they do not require separate sessions or independent checkouts. Shared engineering guidance lives in ordinary docs. Root CLAUDE.md links to AGENTS.md; component adapters import their adjacent instructions. Preserve symlinks when checking out the repository. Find repository skills and optional catalog setup in .agents/README.md. The package index links to component…
We split this agents.md into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 18
Check the supported CLI versions from the repository root.
- Distinctive: line 25
Run the static consistency checks after changing instruction imports, skill adapters, or document locations.
- Distinctive: line 45
Use the target and authorization supplied for the task, and report missing prerequisites.
≈ “Common” is an estimate: at least 1 in 5 of the 36,368 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
Coding agents
You can start at the repository root or in a component directory. Read the applicable AGENTS.md files and follow their links when a change spans components. Component instructions add local constraints; they do not require separate sessions or independent checkouts.
Shared engineering guidance lives in ordinary docs. Root CLAUDE.md links to
AGENTS.md; component adapters import their adjacent instructions. Preserve
symlinks when checking out the repository.
Find repository skills and optional catalog setup in .agents/README.md. The package index links to component guidance.
Verify discovery
Check the supported CLI versions from the repository root. Package owners
verify their existing component discovery separately. In Claude, inspect /context or /memory and the available skills;
in Codex, inspect the loaded instructions and skill list. Confirm the root
instructions, component instructions, and skills for that scope resolve to
their authored files. User/global configuration can add other rules; do not
commit employee settings to compensate for it.
Run the static consistency checks after changing instruction imports, skill adapters, or document locations.
Optional tooling
Repository workflows can use public CLIs, configured integrations, and a user-selected Kubernetes deployment. These are task-specific prerequisites: check-logs needs cluster access only for live collection; supplied logs and source investigation work without it. Codex consultation is optional and uses an available session or configured integration. The retained MCP source and manual sync script are one opt-in configuration. These tools and credentials are not prerequisites for repository instruction discovery.
The Python lint and documentation hooks run locally through the committed Claude adapters. The session-end documentation hook can sync an existing root “Available Skills” table and report undocumented MCP permissions. The current root guide has no such table. The read-only repository checker separately validates links, instructions, and skill adapters.
Use the target and authorization supplied for the task, and report missing prerequisites. Initialize or install catalog skills only when requested. Do not probe unspecified clusters, install global packages, or modify shell configuration.
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 28,693 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 repository27 other files it gives its agents.
- AGENTS.mdAGENTS.md
- io_utils/fsutil/AGENTS.mdAGENTS.md
- kumo/pquery/AGENTS.mdAGENTS.md
- packages/diskgraph/diskgraph-train-table/AGENTS.mdAGENTS.md
- packages/kumo-pql/AGENTS.mdAGENTS.md
- ui/AGENTS.mdAGENTS.md
- io_utils/fsutil/CLAUDE.mdCLAUDE.md
- kumo/pquery/CLAUDE.mdCLAUDE.md
- packages/kumo-ui-catalog/CLAUDE.mdCLAUDE.md
- ui/CLAUDE.mdCLAUDE.md
- kumoai-referenceSkill
- kumo-custom-populationsSkill
- kumo-data-scienceSkill
- kumo-model-explanationsSkill
- kumo-model-persistenceSkill
- kumo-model-tuningSkill
- kumo-training-diagnosticsSkill
- build-uiSkill
- check-logsSkill
- code-uiSkill
- debug-issueSkill
- investigate-training-stageSkill
- preflightSkill
- review-mlSkill
- review-uiSkill
- code-ui-catalogSkill
- review-ui-catalogSkill
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.