arize-skills
AGENTS.md72 starsChanged 37 days agoMITView on GitHub
Skill files (SKILL.md, references/, etc.) are loaded as context for AI agents. Follow these rules when editing them, per the Agent Skills specification: CI (.github/workflows/quality-checks.yml) runs these on every PR. Run the no-Node subset locally with bash scripts/check.sh before pushing:
- Installs packages
In this repository18 other files it gives its agents.
- arize-adminSkill
- arize-ai-provider-integrationSkill
- arize-annotationSkill
- arize-compliance-auditSkill
- arize-datasetSkill
- arize-evaluatorSkill
- arize-experimentSkill
- arize-instrumentation-healthSkill
- arize-instrumentationSkill
- arize-linkSkill
- arize-migrate-braintrustSkill
- arize-migrate-langfuseSkill
- arize-migrate-langsmithSkill
- arize-phoenix-migrationSkill
- arize-prompt-optimizationSkill
- arize-promptsSkill
- arize-span-routingSkill
- arize-traceSkill
The file
Files
Authoring Guidelines for AI Agents
Skill files are read by AI, not humans
Skill files (SKILL.md, references/, etc.) are loaded as context for AI agents. Follow these rules when editing them, per the Agent Skills specification:
- Use Markdown links for every file and doc reference — write
See [the reference guide](references/REFERENCE.md), not a bare path or URL. This is the syntax in the spec's File references example; standardize on it repo-wide for consistency. Keep the target a relative path from the skill root, one level deep (e.g.references/EXAMPLES.md), and avoid deeply nested reference chains. (Runnable commands and code stay literal —scripts/extract.pyin a run instruction,go get …, code blocks — they are commands, not links.) - Additional documentation goes in
references/— supplementary docs (e.g.,ax-profiles.md,ax-setup.md,EXAMPLES.md) belong in thereferences/subdirectory, per the Agent Skills specification. - Keep
SKILL.mdfocused — the full body loads on activation; the spec recommends staying under ~5000 tokens / 500 lines and moving detail intoreferences/.
Quality checks
CI (.github/workflows/quality-checks.yml) runs these on every PR. Run the no-Node subset locally with bash scripts/check.sh before pushing:
scripts/validate_skills.py— per-skill structure: frontmatter,name(kebab-case, 3–64 chars, matches the folder, unique),descriptionlength, declaredassetsexist and are ≤5 MB, Markdown links with relative targets resolve,SKILL.md≤500 lines, and a warning for bare doc paths that should be Markdown links (the rule above).scripts/validate_manifests.py—version.txtmatches every plugin manifest (release-please keeps them in sync; this proves they never drift) and metadata (name,description,keywords,license,repository) is consistent and within limits.- codespell (.codespellrc) and line endings (scripts/check_line_endings.sh) — spelling and LF hygiene.
- Vally lint (eng/vally-lint.mjs) —
@microsoft/vally's static skill lint (spec-compliance + valid references). Needs Node:npm install --no-save @microsoft/vally@0.10.0 && node eng/vally-lint.mjs. - Plugin install smoke test —
claude plugin validate .plus a marketplace add/install of the plugin.
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.

