agentleFS
elizaOSelizaOS/eliza/AGENTS.md

eliza

AGENTS.md20k starsChanged 26 days agoMITView on GitHub

Monorepo for the Eliza agent runtime, application, cloud services, native bridges, benchmarks, and first-party plugins. Read CONTRIBUTING.md and the owning package's README.md before editing; manifests and source are authoritative. This root file is the only repository agent guide. Do not add nested AGENTS.md files. [!IMPORTANT] The contribution rules are required for all issues and PRs. Report real bugs or missing approved MVP requirements. Link the PRD and MVP plan. Do not add features outside that scope. Human maintainers must discuss…

We split this AGENTS.md into the rules it gives an agent

15 rules. About 0 are common, 11 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈11

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Prefer cleanup, removal, and reuse.

    line 25
  2. Distinctive:

    Add new types or code only when necessary.

    line 26
  3. Distinctive:

    Do not create work for minor points with no useful effect, unnecessary tests, defensive code, validation, or truncation.

    line 28
  4. Distinctive:

    Use existing tests. Avoid new unit tests, mock-only proof, and tests that repeat the implementation.

    line 31
  5. Distinctive:

    Every UI PR must include an uploaded MP4 explainer and walkthrough with video evidence of the complete user flow, plus detailed steps to …

    line 33
  6. Distinctive:

    Write all issues and PRs in ASD-STE100 Simplified Technical English.

    line 35
  7. Distinctive:

    Use real LLM calls for evidence.

    line 38
  8. Distinctive:

    Use pinned Bun 1.4.2 and Node 24.15.0, ESM, and the repository's Biome config.

    line 44
  9. Distinctive:

    Keep core independent of hosts.

    line 45
  10. Very specific:

    Use the structured logger and runtime.reportError for runtime diagnostics.

    line 49
  11. Distinctive:

    Keep model-facing context and recorded training data complete.

    line 50
  12. Distinctive:

    Use the existing scheduler, entity/relationship stores, and content-addressed media store.

    line 53
  13. Very specific:

    Keep READMEs and agent guides short.

    line 55
  14. Very specific:

    Use packages/scripts/lib/test-output.ts for stable output paths.

    line 58
  15. Very specific:

    Verify changed behavior with the owning package's tests, typecheck, and lint, then bun run verify.

    line 59

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
AGENTS.md

elizaOS

Monorepo for the Eliza agent runtime, application, cloud services, native bridges, benchmarks, and first-party plugins. Read CONTRIBUTING.md and the owning package's README.md before editing; manifests and source are authoritative. This root file is the only repository agent guide. Do not add nested AGENTS.md files.

[!IMPORTANT] The contribution rules are required for all issues and PRs. Report real bugs or missing approved MVP requirements. Link the PRD and MVP plan. Do not add features outside that scope. Human maintainers must discuss and approve new features, then add them to the PRD and MVP plan before implementation issues or PRs are opened. Maintainers will close unnecessary work and apply contributor penalties.

  • Human-only issue creation: Outside contributors must personally write the issue title and body and submit the issue by hand on the GitHub website. Agents must not write or create issues for contributors. Do not use an API, CLI, browser automation, or another agent to submit an issue. Human approval of an agent-written issue does not meet this rule. An agent that finds a problem must report it privately to the human in the current work session; the human decides whether to write and submit an issue.
  • Prove a useful improvement with before-and-after behavior or relevant scores. Prefer cleanup, removal, and reuse. Combine duplicate types and functions. Add new types or code only when necessary. Explain the research, alternatives, and reason for the chosen implementation.
  • Do not create work for minor points with no useful effect, unnecessary tests, defensive code, validation, or truncation. Preserve required safety boundaries.
  • Validate real behavior end to end and attach results for the reviewed commit. Use existing tests. Avoid new unit tests, mock-only proof, and tests that repeat the implementation. For documentation-only work, check the text and links.
  • Every UI PR must include an uploaded MP4 explainer and walkthrough with video evidence of the complete user flow, plus detailed steps to test the change.
  • Write all issues and PRs in ASD-STE100 Simplified Technical English. Use short, direct sentences, explain technical terms, and write for a non-technical reader.
  • Slop bounty work: star and fork the repository first and work from the fork. Use real LLM calls for evidence. Make no UI changes except broken-flow fixes with before-and-after video. Test in more than one agent client and record each client, version, provider, and model. Rules, environment variables, and the evidence checklist: Slop bounty contributions.
  • Preserve unrelated changes in this shared working tree.
  • Use pinned Bun 1.4.2 and Node 24.15.0, ESM, and the repository's Biome config.
  • Keep core independent of hosts; hosts compose assistant behavior, storage, and model providers. Validate untrusted input at boundaries.
  • Preserve authorization, tenant isolation, cancellation, and effect receipts. Throw typed errors; never fabricate success or hide failures as empty data. Use the structured logger and runtime.reportError for runtime diagnostics.
  • Keep model-facing context and recorded training data complete. Preserve the existing authorized source-selection/restoration contract; do not add silent truncation, summaries, or recency limits. Reject hard size limits explicitly.
  • Use the existing scheduler, entity/relationship stores, and content-addressed media store. Do not introduce parallel ownership or bypass SSRF guards.
  • Keep READMEs and agent guides short. Retain other Markdown only when consumed by runtime, build, tests, publishing, or required attribution. Generated reports belong in ignored repository-root test-results/, with one leaf per producer. Use packages/scripts/lib/test-output.ts for stable output paths.
  • Verify changed behavior with the owning package's tests, typecheck, and lint, then bun run verify. Validate documentation links for docs-only changes. UI changes require the app visual audit and desktop/mobile inspection.
  • Submit changes through a PR against develop. Report vulnerabilities privately through GitHub Security Advisories. The community plugin registry is retired; first-party catalog data lives in packages/core/src/catalog/.

Getting started, build, test, and benchmark commands: README.md.

Design inventories are advisory; do not gate builds on saved component counts, story percentages, or expired review metadata.

Download now (ZIP)

Trending

Steady: 0 new projects took it in the last 4 weeks, none the 4 weeks before.thin data

Week of Aug 30: 0 new projectsWeek of Sep 6: 0 new projectsWeek of Sep 13: 0 new projectsWeek of Sep 20: 0 new projectsWeek of Sep 27: 0 new projectsWeek of Oct 4: 0 new projectsWeek of Oct 11: 0 new projectsWeek of Oct 18: 0 new projects
Aug 2026Sep 2026Oct 2026
New projects taking a version of this file each week, over the 8 weeks our latest run read: only 0 in all, so this is a hint, not a trend.

Who's talking about elizaOS/eliza

The biggest 5 of 66 mentions linking to the repository that our crawl found. Each line is their own headline.

See all 66

Talked about on GitHub, Bluesky, Hacker News and the web (66 times).

On GitHub · 50

45 more on GitHub

On Bluesky · 12

7 more on Bluesky

On the web · 1

  • 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.

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

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.