{"id":50989603,"url":"https://github.com/jcurbelo/agent-seed","last_synced_at":"2026-06-20T00:30:55.283Z","repository":{"id":359118637,"uuid":"1178131513","full_name":"jcurbelo/agent-seed","owner":"jcurbelo","description":"Self-evolving AI assistant starter kit. Markdown-only. Works with Claude Code, Codex, Devin, Cursor, and any LLM agent","archived":false,"fork":false,"pushed_at":"2026-05-20T12:40:58.000Z","size":36,"stargazers_count":6,"open_issues_count":0,"forks_count":1,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-05-20T17:12:37.140Z","etag":null,"topics":["ai-agent","claude-code","codex","cursor","devin","llm","markdown","personal-assistant","self-improving","starter-kit"],"latest_commit_sha":null,"homepage":"","language":"Shell","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/jcurbelo.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":"2026-03-10T18:07:59.000Z","updated_at":"2026-05-20T12:41:03.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/jcurbelo/agent-seed","commit_stats":null,"previous_names":["jcurbelo/agent-seed"],"tags_count":null,"template":true,"template_full_name":null,"purl":"pkg:github/jcurbelo/agent-seed","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/jcurbelo%2Fagent-seed","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/jcurbelo%2Fagent-seed/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/jcurbelo%2Fagent-seed/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/jcurbelo%2Fagent-seed/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/jcurbelo","download_url":"https://codeload.github.com/jcurbelo/agent-seed/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/jcurbelo%2Fagent-seed/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":34553456,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-06-19T02:00:06.005Z","response_time":61,"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-agent","claude-code","codex","cursor","devin","llm","markdown","personal-assistant","self-improving","starter-kit"],"created_at":"2026-06-20T00:30:53.772Z","updated_at":"2026-06-20T00:30:55.273Z","avatar_url":"https://github.com/jcurbelo.png","language":"Shell","funding_links":[],"categories":[],"sub_categories":[],"readme":"# agent-seed\n\n**Self-evolving AI assistant starter kit. Markdown-only. Works with any LLM agent.**\n\nPlant the seed. It grows with you.\n\n---\n\n## What is this?\n\nA collection of markdown files that teach AI agents who you are — your voice, your preferences, your work context, your priorities.\n\nIt works with [Claude Code](https://docs.anthropic.com/en/docs/claude-code), [Codex](https://openai.com/index/introducing-codex/), [Devin](https://devin.ai), [Cursor](https://cursor.sh), [Windsurf](https://codeium.com/windsurf), and anything else that reads markdown.\n\nNo API keys. No runtime. No vendor lock-in. Just `.md` files.\n\n### What it does\n\n- **Gives any AI your voice** — communication style, tone, patterns\n- **Tracks your daily priorities** — todos, drafts, notes\n- **Self-improves** — learns from each session, proposes new agents via PRs\n- **Stays small** — enforced size limits prevent markdown bloat\n- **Grows with you** — `/setup` generates agents based on YOUR workflow, `/learn` evolves them over time\n\n### What it doesn't do\n\n- Doesn't require API keys or expose secrets\n- Doesn't lock you into any tool or vendor\n- Doesn't run any code — it's just context files\n\n## Quick Start\n\n### 1. Use this template\n\nClick **\"Use this template\"** on GitHub, or:\n\n```bash\ngit clone https://github.com/jcurbelo/agent-seed.git my-agent\ncd my-agent\nrm -rf .git \u0026\u0026 git init\n```\n\n### 2. Run setup\n\nOpen the repo in your preferred AI coding tool and run:\n\n```\n/setup\n```\n\nThe setup wizard will ask about your name, role, communication style, and tools — then generate personalized agents, voice profiles, and context files.\n\n### 3. Start using it\n\n```\n/standup        # Morning briefing — scan todos, plan the day\n/review         # Quick check-in — what needs attention?\n/eod            # End of day — capture learnings, prep tomorrow\n/learn          # Self-improvement — prune, evolve, grow new agents\n/handoff        # Compact the session into a handoff doc for the next agent\n/grill-me       # Stress-test a plan; resolve every branch of the decision tree\n/caveman        # Ultra-compressed reply mode (~75% fewer tokens)\n/write-a-skill  # Author a new skill with the right frontmatter + conventions\n```\n\n## How It Works\n\n### The Index Pattern\n\nEvery folder has an `index.md` that links to detail pages — like imports in code:\n\n```\ncontext/\n├── index.md              # Master index → links to voice/, work/\n├── voice/\n│   ├── index.md          # Voice overview → links to patterns.md\n│   └── patterns.md       # Detailed examples\n└── work/\n    ├── index.md          # Role overview → links to tools.md\n    └── tools.md          # Services and repos you use\n```\n\nIndex files are always loaded first. Detail pages load on demand. Size limits keep everything fast.\n\n### Self-Evolution\n\nThis is the core idea. Three layers:\n\n1. **Reflection agent** — after meaningful sessions, captures learnings to `memory/`\n2. **`/learn` skill** — periodic self-improvement: prunes stale entries, identifies gaps, creates new agents via PR\n3. **`/setup` re-run** — add new integrations or agents anytime\n\nWhen the system detects a gap (e.g., you keep asking about Slack but there's no Slack agent), it creates one and files a PR.\n\n### Frontmatter\n\nEvery `.md` file has YAML frontmatter that defines its role:\n\n```yaml\n# Agent definition\n---\nmodel: sonnet                    # Model tier: haiku | sonnet | opus\ntools: [Read, Write, Edit]      # Allowed tools\npermissionMode: default          # default | bypassPermissions\nmaxTurns: 10                     # Max agent turns\n---\n\n# Scoped rule\n---\npaths: [\"todos/**\"]              # When these paths are active\n---\n\n# Context/memory page\n---\ntitle: Voice Profile\ndescription: Communication style and tone\nparent: ../index.md              # Navigation link\nmax_lines: 250                   # Enforced by policy script\n---\n```\n\n### Markdown Policy\n\nA script (`scripts/check-markdown-policy.sh`) enforces:\n\n- Size limits per file type (index: 150 lines, pages: 250 lines, memory: 200 lines)\n- Required frontmatter keys\n- Required todo sections\n- No orphaned pages (every page linked from an index)\n\nRun it locally or add it to CI.\n\n## File Structure\n\n```\nagent-seed/\n├── CLAUDE.md                     # Claude Code / LLM entrypoint\n├── AGENTS.md                     # Devin / Codex entrypoint\n├── .cursorrules                  # Cursor entrypoint\n├── README.md                     # You are here\n├── LICENSE                       # MIT\n│\n├── .claude/\n│   ├── settings.json             # Safety hooks\n│   ├── agents/                   # Agent definitions\n│   │   ├── index.md              # Agent registry + routing\n│   │   ├── researcher.md         # Web research\n│   │   ├── coder.md              # Code writing\n│   │   ├── reviewer.md           # Code review\n│   │   ├── todo-manager.md       # Daily todos\n│   │   └── reflection.md         # Self-improvement engine\n│   ├── skills/                   # Slash commands\n│   │   ├── setup/SKILL.md         # /setup wizard\n│   │   ├── standup/SKILL.md       # Morning briefing\n│   │   ├── eod/SKILL.md           # End of day\n│   │   ├── review/SKILL.md        # Quick check-in\n│   │   ├── learn/SKILL.md         # Self-improvement\n│   │   ├── update-todos/SKILL.md  # Interactive todo review\n│   │   ├── handoff/SKILL.md       # Session handoff doc (mattpocock)\n│   │   ├── grill-me/SKILL.md      # Plan stress-test (mattpocock)\n│   │   ├── caveman/SKILL.md       # Ultra-compressed replies (mattpocock)\n│   │   └── write-a-skill/SKILL.md # New-skill authoring (mattpocock)\n│   └── rules/                    # Scoped rules (load by file path)\n│       ├── todos.md\n│       ├── context.md\n│       └── drafts.md\n│\n├── context/                      # Who you are (populated by /setup)\n│   ├── index.md\n│   ├── voice/\n│   │   ├── index.md\n│   │   └── patterns.md\n│   └── work/\n│       ├── index.md\n│       └── tools.md\n│\n├── memory/                       # What the AI has learned\n│   ├── MEMORY.md                 # Master index (always loaded)\n│   ├── learnings.md\n│   ├── preferences.md\n│   ├── patterns.md\n│   ├── improvements.md\n│   └── external-skill-sources.md # Attribution for imported skills\n│\n├── todos/                        # Daily task tracking\n│   └── _template.md\n│\n├── handoffs/                     # Session handoff docs (committed, never /tmp)\n│   └── YYYY-MM-DD-\u003cslug\u003e.md\n│\n└── scripts/\n    └── check-markdown-policy.sh  # Enforces structure + size limits\n```\n\n## Attribution\n\nSome skills are adapted from external sources. See [`memory/external-skill-sources.md`](memory/external-skill-sources.md) for full attribution and source links.\n\n## Compatibility\n\n| Tool | Entrypoint | Status |\n|------|-----------|--------|\n| Claude Code | `CLAUDE.md` | Full support (agents, skills, rules, hooks) |\n| Cursor | `.cursorrules` | Context + rules |\n| Codex | `AGENTS.md` | Context + agents |\n| Devin | `AGENTS.md` | Context + agents |\n| Windsurf | `CLAUDE.md` | Context + rules |\n| Any LLM | `CLAUDE.md` | Read as context |\n\n## Philosophy\n\nYour AI assistant should know you the way a good colleague does — your communication style, your priorities, your tools. This repo is that knowledge, version-controlled and self-evolving.\n\nUse it with current AI coding agents. Don't expose private keys. Do whatever you want with it.\n\n## License\n\nMIT — do whatever you want.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjcurbelo%2Fagent-seed","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fjcurbelo%2Fagent-seed","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjcurbelo%2Fagent-seed/lists"}