{"id":39436880,"url":"https://github.com/leochiu-a/universal-agents","last_synced_at":"2026-01-18T04:14:39.261Z","repository":{"id":324186105,"uuid":"1096269609","full_name":"leochiu-a/universal-agents","owner":"leochiu-a","description":"The Lightest Shared Standard for AI Agents","archived":false,"fork":false,"pushed_at":"2025-12-05T10:03:09.000Z","size":872,"stargazers_count":27,"open_issues_count":1,"forks_count":2,"subscribers_count":0,"default_branch":"main","last_synced_at":"2025-12-08T18:51:50.558Z","etag":null,"topics":["ai","ai-agents","claude-code","codex","gemini-cli"],"latest_commit_sha":null,"homepage":"https://leochiu-a.github.io/universal-agents/","language":"TypeScript","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/leochiu-a.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":"AGENTS.md","dco":null,"cla":null}},"created_at":"2025-11-14T07:16:40.000Z","updated_at":"2025-12-06T07:47:00.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/leochiu-a/universal-agents","commit_stats":null,"previous_names":["leochiu-a/universal-agents"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/leochiu-a/universal-agents","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/leochiu-a%2Funiversal-agents","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/leochiu-a%2Funiversal-agents/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/leochiu-a%2Funiversal-agents/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/leochiu-a%2Funiversal-agents/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/leochiu-a","download_url":"https://codeload.github.com/leochiu-a/universal-agents/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/leochiu-a%2Funiversal-agents/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":28529465,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-01-18T00:39:45.795Z","status":"online","status_checked_at":"2026-01-18T02:00:07.578Z","response_time":98,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["ai","ai-agents","claude-code","codex","gemini-cli"],"created_at":"2026-01-18T04:14:38.713Z","updated_at":"2026-01-18T04:14:39.252Z","avatar_url":"https://github.com/leochiu-a.png","language":"TypeScript","readme":"# Universal Agents\n\nA lightweight standard for sharing agent configurations across all AI coding assistants using a single `AGENTS.md` file.\n\n![](public/image.png)\n\n## Quick Start\n\nRun the CLI via npx:\n\n```bash\nnpx universal-agents init           # copy AGENTS.md into your project\nnpx universal-agents create skill   # scaffold .agents/skills/\u003cname\u003e/SKILL.md\nnpx universal-agents create rule    # scaffold .agents/rules/\u003cname\u003e.md\n```\n\n### Apply Universal Agents in Claude Code\n\nClaude Code sometimes ignores its `CLAUDE.md` file and doesn't support the `AGENTS.md` protocol. Use a user memory prompt to force loading:\n\n```\nALWAYS read AGENTS.md file first\n```\n\nIn Claude, pin this via the `/memory` command so it stays in user memory.\n\n## Why Universal Agents?\n\nDifferent AI coding assistants use different conventions to organize their configuration files. For example:\n\n- Claude Code stores skills in `.claude/skills/\u003cskill-name\u003e/SKILL.md`\n- Cursor stores rules in the `.cursor/rules/` folder\n\nThis fragmentation locks configurations into single platforms. Teams cannot share context and workflow definitions across tools.\n\n## The Solution\n\nUniversal Agents provides the **lightest, simplest** shared standard: `AGENTS.md` + `.agents/` folder. Any agent—regardless of runtime—reads this one manifest to bootstrap rules and skills.\n\n## How it works\n\nUniversal Agents uses a single entry point—`AGENTS.md`—to describe which skills and rules apply to any task. All skills and rules live in `.agents/` for version control and easy sharing.\n\n| File/Folder       | Purpose                                                                                                       |\n| ----------------- | ------------------------------------------------------------------------------------------------------------- |\n| `AGENTS.md`       | Control manifest listing all available skills and rules, with trigger conditions and loading instructions.    |\n| `.agents/skills/` | Reusable task workflows (e.g., code review, testing) — each skill is self-contained and production-ready.     |\n| `.agents/rules/`  | Domain guidelines (e.g., API design, React conventions) — long-lived constraints that enforce team standards. |\n\n## Key advantages\n\n- **One source of truth**: `AGENTS.md` works across Claude Code, Cursor, Codex, and any future AI assistant.\n- **Zero boilerplate**: No tool-specific file structure. Just read `AGENTS.md`, follow its protocol.\n- **Minimal footprint**: Lightweight Markdown—no DSLs, config files, or complex tooling required.\n- **Immediate adoption**: Agents that understand the `AGENTS.md` protocol can bootstrap any project instantly.\n\n## Execution protocol\n\n1. **Read `AGENTS.md` first** to discover available skills and rules, plus their trigger conditions.\n2. **Load skills on demand**: when a task matches a skill's trigger, read the skill file and follow its workflow.\n3. **Enforce rules whenever relevant**: if a task touches governed domains, preload and apply the matching rule file.\n4. **Declare applied guardrails**: list which skills and rules were active so transparency is guaranteed.\n\n## Visual references\n\n| Asset   | Codex                              | Claude code                                 |\n| ------- | ---------------------------------- | ------------------------------------------- |\n| Preview | ![Codex diagram](public/codex.png) | ![Claude code view](public/claude-code.png) |\n\n## Rules vs. skills in plain language\n\n| Aspect           | Rules (Cursor-style)                                                               | Skills (Claude-style)                                                                                            |\n| ---------------- | ---------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------- |\n| Content type     | Textual guardrails (e.g., “use `python-docx`, add a title, keep formatting tidy”). | End-to-end package with docs, runnable templates, and tooling (often sourced from `/mnt/skills/.../SKILL.md`).   |\n| Execution effort | Agent interprets and implements instructions manually.                             | Provides production-tested code snippets, patterns, and edge-case playbooks ready to drop into a task.           |\n| Use case         | Keeps behavior aligned with policies and coding standards.                         | Speeds up specialized workflows by handing the agent a toolbox plus demo-quality walkthroughs.                   |\n| Maintenance      | Update the prose rule whenever governance changes.                                 | Refresh the skill bundle when better implementations or examples land; the structure makes verification simpler. |\n\nIn short:\n\n- Rules are the \"manual\"\n- Skills are the \"manual + toolbox + show-and-tell\" kit you reach for when the task demands more than guidelines.\n","funding_links":[],"categories":[],"sub_categories":[],"project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fleochiu-a%2Funiversal-agents","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fleochiu-a%2Funiversal-agents","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fleochiu-a%2Funiversal-agents/lists"}