1 line
95 KiB
JSON
1 line
95 KiB
JSON
{"repo": "carltonkenny/genesis", "files": {"README.md": {"8a806f6470ce5a5a488e6f68d5fb2a5240fafe21": {"t": 1785746604.991209, "n": 9517, "h": "d73f89554e6edbb5", "c": "# GenesisX\n\n> **Every project begins here.** GenesisX reads your entire codebase and builds a tailored team of AI agents — each owning one domain, carrying real architecture knowledge, and scoped with file-level permissions.\n\n> **v0.2** — Weighted sampling, persistent memory (.genesis/state.json), dependency graph, kanban board, agent validation. 6 hardening features shipped. Agent output quality: 8-9/10.\n\n```bash\nnpm install -g genesisx\ncd any-project\ngenesis .\n```\n\n---\n\n## Real Results — 5 GitHub Repositories Tested\n\n### fastapi (Python, 2935 files)\n\n```\n┌──────────────────────────────────────────────┐\n│ fastapi · Python · 8 files │\n│ DeepSeek V3 · ~$0.002 · 5 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ fastapi/ │\n│ tests/ │\n│ docs/ docs_src/ │\n│ scripts/ │\n│ fastapi-slim/ │\n│ 8 bugs · 14 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ fastapi-core-engineer fastapi/ 49 files│\n│ fastapi-test-engineer tests/ 581│\n│ documentation-agent docs/ 2235│\n│ scripts-maintainer scripts/ 69│\n│ fastapi-slim-maintainer fastapi-slim/ 1│\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\nGenerated agent: `fastapi-core-engineer.md` (85 lines). Knows the exact FastAPI version (0.136.3), every core module, all import conventions, async patterns, test framework, and deprecation strategy — extracted straight from the source.\n\n### gin (Go, 53 files) — 2 bugs found\n\n```\n┌──────────────────────────────────────────────┐\n│ gin · Go · 53 files │\n│ DeepSeek V3 · ~$0.002 · 5 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ binding/ render/ internal/ ginS/ │\n│ codec/ docs/ examples/ testdata/ │\n│ 2 bugs · 4 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ gin-core-engineer core modules 54 │\n│ codec-dev codec/ 5 │\n│ docs-agent docs/ 1 │\n│ examples-maintainer examples/ 1 │\n│ testdata-domain-agent testdata/ 7 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n### express (TypeScript, 10 entries) — 1 bug found\n\n```\n┌──────────────────────────────────────────────┐\n│ express · TypeScript (Express) · 10 files │\n│ DeepSeek V3 · ~$0.002 · 3 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ lib/ examples/ test/ │\n│ 1 bug · 8 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ express-core-engineer lib/ 6 │\n│ express-examples-agent examples/ 80 │\n│ express-test-engineer test/ 112 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n### click (Python, 9 entries) — clean codebase\n\n```\n┌──────────────────────────────────────────────┐\n│ click · Python · 9 files │\n│ DeepSeek V3 · ~$0.002 · 4 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ src/ docs/ examples/ tests/ │\n│ 0 bugs · 1 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ core-dev src/ 18 │\n│ click-documentation-agent docs/ 41 │\n│ click-examples-auditor examples/ 39 │\n│ click-test-engineer tests/ 31 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n### got (TypeScript, 9 entries) — 3 bugs + 1 secret\n\n```\n┌──────────────────────────────────────────────┐\n│ got · TypeScript (Express) · 9 files │\n│ DeepSeek V3 · ~$0.002 · 5 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ source/ test/ documentation/ │\n│ media/ benchmark/ │\n│ 3 bugs · 1 secret · 2 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ got-source-agent source/ 23 │\n│ got-test-domain-agent test/ 49 │\n│ documentation-agent documentation/ 26 │\n│ media-asset-manager media/ 4 │\n│ benchmark-agent benchmark/ 2 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n---\n\n## Quick Start\n\n```bash\n# 1. Install\nnpm install -g genesisx\n\n# 2. Set your LLM key (any one works)\nexport DEEPSEEK_API_KEY=sk-... # recommended: cheapest, $0.002/run\nexport ANTHROPIC_API_KEY=sk-ant-... # Claude\nexport OPENAI_API_KEY=sk-... # GPT-4o\n\n# 3. Run on any project\ncd any-project\ngenesis .\n\n# Or skip approval for CI/automation\ngenesis . --yes\ngenesis . --provider claude\ngenesis . --dry-run # preview without writing\ngenesis . --json # machine-readable output\n```\n\n---\n\n## What Gets Created\n\n```\nyour-project/\n├── prompts/\n│ ├── backend-api-engineer.md # Agent owning backend/ · 36 files\n│ ├── frontend-crafter.md # Agent owning frontend/ · 12 files\n│ ├── db-engineer.md # Agent owning database/ · 8 files\n│ └── reviewer.md # Read-only reviewer · gates all merges\n├── PROJECT-CONTEXT.md # Full AI narrative of your codebase\n└── opencode.json # Registered agent team\n```\n\n---\n\n## How It Works\n\n```\nPhase 1: Survey\nReads directory tree, tech stack, dependencies, imports, configs, tests.\n\nPhase 2: Pre-Analyze (local, 0 tokens, 0 cost)\nDetects hardcoded secrets, AI slop, long files, circular imports,\nmissing error handling, untested files, dead code.\n\nPhase 3: Domain Detection (LLM, ~$0.001)\nSends directory structure + quality flags. LLM identifies natural\ndomain boundaries — the seams in your codebase.\n\nPhase 4: Deep Analysis (LLM, ~$0.005/domain)\nFor small/medium domains, sends every file. For large domains, uses weighted sampling\n(imports + bugs + boundary proximity) to fit context while staying high-signal.\nFinds bugs at file:line precision. Detects AI-generated code. Flags security issues.\nGenerates the agent prompt with knowledge extracted from real code.\n\nPhase 5: Build (local, 0 tokens)\nRenders agent .md files, writes PROJECT-CONTEXT.md, registers team\nin opencode.json.\n\nApproval: You review findings before anything is written. Default yes.\n```\n\n---\n\n## Providers\n\n| Provider | Model | Cost/Run* | Context | Setup |\n|----------|-------|-----------|---------|-------|\n| **DeepSeek** | V3 | $0.002 | 128K | `export DEEPSEEK_API_KEY=sk-...` |\n| Claude | Sonnet 4 | $0.08 | 200K | `export ANTHROPIC_API_KEY=sk-ant-...` |\n| OpenAI | GPT-4o | $0.05 | 128K | `export OPENAI_API_KEY=sk-...` |\n\n*For a 120-file project. Genesis auto-detects which provider to use from your environment.\n\n[Full cost breakdown →](docs/TOKEN-COSTS.md)\n\n---\n\n## MCP Server\n\nGenesis also runs as an MCP server — connect it to OpenCode, Claude Code, or Cursor:\n\n```json\n{\n \"mcpServers\": {\n \"genesis\": {\n \"command\": \"npx\",\n \"args\": [\"genesis-mcp\"]\n }\n }\n}\n```\n\nTools: `survey_project`, `propose_team`, `build_team`. Run Genesis from inside your editor — no terminal needed.\n\n---\n\n## Principles\n\n- **Every file is scanned locally.** Deep analysis uses weighted sampling for large domains to fit context.\n- **No token limits.** Choose DeepSeek and it's effectively free (~500 runs per $1).\n- **Your code stays local.** Sent directly to the API you choose. Never cached. Never stored.\n- **Reviewer always included.** Every team gets a read-only gate. Non-negotiable.\n- **File-scoped permissions.** Agents can only edit their domain. Real filesystem paths.\n\n---\n\n## Docs\n\n| Doc | What |\n|-----|------|\n| [PRD](docs/PRD.md) | What, why, who, success metrics |\n| [Architecture](docs/ARCHITECTURE.md) | System design, data flow, provider model |\n| [Rules](docs/RULES.md) | How Genesis builds itself using its own agents |\n| [Token Costs](docs/TOKEN-COSTS.md) | Per-provider cost table, no-limit strategy |\n| [Roadmap](docs/ROADMAP.md) | 14-day plan, known blockers, mitigations |\n| [Scenarios](docs/SCENARIOS.md) | 5 real-world use cases with terminal output |\n| [Difficulties](docs/DIFFICULTIES.md) | What will break and how we handle it |\n\n---\n\nMIT License · GenesisX v0.2\n"}, "e7a6ee8bc2ad493f0e7614bc0a1bd9ebcb01e362": {"t": 1785766227.3183973, "n": 9517, "h": "d73f89554e6edbb5", "c": "# GenesisX\n\n> **Every project begins here.** GenesisX reads your entire codebase and builds a tailored team of AI agents — each owning one domain, carrying real architecture knowledge, and scoped with file-level permissions.\n\n> **v0.2** — Weighted sampling, persistent memory (.genesis/state.json), dependency graph, kanban board, agent validation. 6 hardening features shipped. Agent output quality: 8-9/10.\n\n```bash\nnpm install -g genesisx\ncd any-project\ngenesis .\n```\n\n---\n\n## Real Results — 5 GitHub Repositories Tested\n\n### fastapi (Python, 2935 files)\n\n```\n┌──────────────────────────────────────────────┐\n│ fastapi · Python · 8 files │\n│ DeepSeek V3 · ~$0.002 · 5 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ fastapi/ │\n│ tests/ │\n│ docs/ docs_src/ │\n│ scripts/ │\n│ fastapi-slim/ │\n│ 8 bugs · 14 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ fastapi-core-engineer fastapi/ 49 files│\n│ fastapi-test-engineer tests/ 581│\n│ documentation-agent docs/ 2235│\n│ scripts-maintainer scripts/ 69│\n│ fastapi-slim-maintainer fastapi-slim/ 1│\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\nGenerated agent: `fastapi-core-engineer.md` (85 lines). Knows the exact FastAPI version (0.136.3), every core module, all import conventions, async patterns, test framework, and deprecation strategy — extracted straight from the source.\n\n### gin (Go, 53 files) — 2 bugs found\n\n```\n┌──────────────────────────────────────────────┐\n│ gin · Go · 53 files │\n│ DeepSeek V3 · ~$0.002 · 5 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ binding/ render/ internal/ ginS/ │\n│ codec/ docs/ examples/ testdata/ │\n│ 2 bugs · 4 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ gin-core-engineer core modules 54 │\n│ codec-dev codec/ 5 │\n│ docs-agent docs/ 1 │\n│ examples-maintainer examples/ 1 │\n│ testdata-domain-agent testdata/ 7 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n### express (TypeScript, 10 entries) — 1 bug found\n\n```\n┌──────────────────────────────────────────────┐\n│ express · TypeScript (Express) · 10 files │\n│ DeepSeek V3 · ~$0.002 · 3 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ lib/ examples/ test/ │\n│ 1 bug · 8 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ express-core-engineer lib/ 6 │\n│ express-examples-agent examples/ 80 │\n│ express-test-engineer test/ 112 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n### click (Python, 9 entries) — clean codebase\n\n```\n┌──────────────────────────────────────────────┐\n│ click · Python · 9 files │\n│ DeepSeek V3 · ~$0.002 · 4 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ src/ docs/ examples/ tests/ │\n│ 0 bugs · 1 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ core-dev src/ 18 │\n│ click-documentation-agent docs/ 41 │\n│ click-examples-auditor examples/ 39 │\n│ click-test-engineer tests/ 31 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n### got (TypeScript, 9 entries) — 3 bugs + 1 secret\n\n```\n┌──────────────────────────────────────────────┐\n│ got · TypeScript (Express) · 9 files │\n│ DeepSeek V3 · ~$0.002 · 5 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ source/ test/ documentation/ │\n│ media/ benchmark/ │\n│ 3 bugs · 1 secret · 2 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ got-source-agent source/ 23 │\n│ got-test-domain-agent test/ 49 │\n│ documentation-agent documentation/ 26 │\n│ media-asset-manager media/ 4 │\n│ benchmark-agent benchmark/ 2 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n---\n\n## Quick Start\n\n```bash\n# 1. Install\nnpm install -g genesisx\n\n# 2. Set your LLM key (any one works)\nexport DEEPSEEK_API_KEY=sk-... # recommended: cheapest, $0.002/run\nexport ANTHROPIC_API_KEY=sk-ant-... # Claude\nexport OPENAI_API_KEY=sk-... # GPT-4o\n\n# 3. Run on any project\ncd any-project\ngenesis .\n\n# Or skip approval for CI/automation\ngenesis . --yes\ngenesis . --provider claude\ngenesis . --dry-run # preview without writing\ngenesis . --json # machine-readable output\n```\n\n---\n\n## What Gets Created\n\n```\nyour-project/\n├── prompts/\n│ ├── backend-api-engineer.md # Agent owning backend/ · 36 files\n│ ├── frontend-crafter.md # Agent owning frontend/ · 12 files\n│ ├── db-engineer.md # Agent owning database/ · 8 files\n│ └── reviewer.md # Read-only reviewer · gates all merges\n├── PROJECT-CONTEXT.md # Full AI narrative of your codebase\n└── opencode.json # Registered agent team\n```\n\n---\n\n## How It Works\n\n```\nPhase 1: Survey\nReads directory tree, tech stack, dependencies, imports, configs, tests.\n\nPhase 2: Pre-Analyze (local, 0 tokens, 0 cost)\nDetects hardcoded secrets, AI slop, long files, circular imports,\nmissing error handling, untested files, dead code.\n\nPhase 3: Domain Detection (LLM, ~$0.001)\nSends directory structure + quality flags. LLM identifies natural\ndomain boundaries — the seams in your codebase.\n\nPhase 4: Deep Analysis (LLM, ~$0.005/domain)\nFor small/medium domains, sends every file. For large domains, uses weighted sampling\n(imports + bugs + boundary proximity) to fit context while staying high-signal.\nFinds bugs at file:line precision. Detects AI-generated code. Flags security issues.\nGenerates the agent prompt with knowledge extracted from real code.\n\nPhase 5: Build (local, 0 tokens)\nRenders agent .md files, writes PROJECT-CONTEXT.md, registers team\nin opencode.json.\n\nApproval: You review findings before anything is written. Default yes.\n```\n\n---\n\n## Providers\n\n| Provider | Model | Cost/Run* | Context | Setup |\n|----------|-------|-----------|---------|-------|\n| **DeepSeek** | V3 | $0.002 | 128K | `export DEEPSEEK_API_KEY=sk-...` |\n| Claude | Sonnet 4 | $0.08 | 200K | `export ANTHROPIC_API_KEY=sk-ant-...` |\n| OpenAI | GPT-4o | $0.05 | 128K | `export OPENAI_API_KEY=sk-...` |\n\n*For a 120-file project. Genesis auto-detects which provider to use from your environment.\n\n[Full cost breakdown →](docs/TOKEN-COSTS.md)\n\n---\n\n## MCP Server\n\nGenesis also runs as an MCP server — connect it to OpenCode, Claude Code, or Cursor:\n\n```json\n{\n \"mcpServers\": {\n \"genesis\": {\n \"command\": \"npx\",\n \"args\": [\"genesis-mcp\"]\n }\n }\n}\n```\n\nTools: `survey_project`, `propose_team`, `build_team`. Run Genesis from inside your editor — no terminal needed.\n\n---\n\n## Principles\n\n- **Every file is scanned locally.** Deep analysis uses weighted sampling for large domains to fit context.\n- **No token limits.** Choose DeepSeek and it's effectively free (~500 runs per $1).\n- **Your code stays local.** Sent directly to the API you choose. Never cached. Never stored.\n- **Reviewer always included.** Every team gets a read-only gate. Non-negotiable.\n- **File-scoped permissions.** Agents can only edit their domain. Real filesystem paths.\n\n---\n\n## Docs\n\n| Doc | What |\n|-----|------|\n| [PRD](docs/PRD.md) | What, why, who, success metrics |\n| [Architecture](docs/ARCHITECTURE.md) | System design, data flow, provider model |\n| [Rules](docs/RULES.md) | How Genesis builds itself using its own agents |\n| [Token Costs](docs/TOKEN-COSTS.md) | Per-provider cost table, no-limit strategy |\n| [Roadmap](docs/ROADMAP.md) | 14-day plan, known blockers, mitigations |\n| [Scenarios](docs/SCENARIOS.md) | 5 real-world use cases with terminal output |\n| [Difficulties](docs/DIFFICULTIES.md) | What will break and how we handle it |\n\n---\n\nMIT License · GenesisX v0.2\n"}, "dfec0049f7d5daa333bdb86903070dc8de9f6155": {"t": 1785766229.120157, "n": 9395, "h": "a76d5d40887ce338", "c": "# GenesisX\n\n> **Every project begins here.** GenesisX reads your entire codebase and builds a tailored team of AI agents — each owning one domain, carrying real architecture knowledge, and scoped with file-level permissions.\n\n> **v0.2** — Weighted sampling, persistent memory (.genesis/state.json), dependency graph, kanban board, agent validation. 6 hardening features shipped. Agent output quality: 8-9/10.\n\n```bash\nnpm install -g genesisx\ncd any-project\ngenesis .\n```\n\n---\n\n## Real Results — 5 GitHub Repositories Tested\n\n### fastapi (Python, 2935 files)\n\n```\n┌──────────────────────────────────────────────┐\n│ fastapi · Python · 8 files │\n│ DeepSeek V3 · ~$0.002 · 5 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ fastapi/ │\n│ tests/ │\n│ docs/ docs_src/ │\n│ scripts/ │\n│ fastapi-slim/ │\n│ 8 bugs · 14 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ fastapi-core-engineer fastapi/ 49 files│\n│ fastapi-test-engineer tests/ 581│\n│ documentation-agent docs/ 2235│\n│ scripts-maintainer scripts/ 69│\n│ fastapi-slim-maintainer fastapi-slim/ 1│\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\nGenerated agent: `fastapi-core-engineer.md` (85 lines). Knows the exact FastAPI version (0.136.3), every core module, all import conventions, async patterns, test framework, and deprecation strategy — extracted straight from the source.\n\n### gin (Go, 53 files) — 2 bugs found\n\n```\n┌──────────────────────────────────────────────┐\n│ gin · Go · 53 files │\n│ DeepSeek V3 · ~$0.002 · 5 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ binding/ render/ internal/ ginS/ │\n│ codec/ docs/ examples/ testdata/ │\n│ 2 bugs · 4 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ gin-core-engineer core modules 54 │\n│ codec-dev codec/ 5 │\n│ docs-agent docs/ 1 │\n│ examples-maintainer examples/ 1 │\n│ testdata-domain-agent testdata/ 7 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n### express (TypeScript, 10 entries) — 1 bug found\n\n```\n┌──────────────────────────────────────────────┐\n│ express · TypeScript (Express) · 10 files │\n│ DeepSeek V3 · ~$0.002 · 3 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ lib/ examples/ test/ │\n│ 1 bug · 8 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ express-core-engineer lib/ 6 │\n│ express-examples-agent examples/ 80 │\n│ express-test-engineer test/ 112 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n### click (Python, 9 entries) — clean codebase\n\n```\n┌──────────────────────────────────────────────┐\n│ click · Python · 9 files │\n│ DeepSeek V3 · ~$0.002 · 4 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ src/ docs/ examples/ tests/ │\n│ 0 bugs · 1 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ core-dev src/ 18 │\n│ click-documentation-agent docs/ 41 │\n│ click-examples-auditor examples/ 39 │\n│ click-test-engineer tests/ 31 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n### got (TypeScript, 9 entries) — 3 bugs + 1 secret\n\n```\n┌──────────────────────────────────────────────┐\n│ got · TypeScript (Express) · 9 files │\n│ DeepSeek V3 · ~$0.002 · 5 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ source/ test/ documentation/ │\n│ media/ benchmark/ │\n│ 3 bugs · 1 secret · 2 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ got-source-agent source/ 23 │\n│ got-test-domain-agent test/ 49 │\n│ documentation-agent documentation/ 26 │\n│ media-asset-manager media/ 4 │\n│ benchmark-agent benchmark/ 2 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n---\n\n## Quick Start\n\n```bash\n# 1. Install\nnpm install -g genesisx\n\n# 2. Set your LLM key (any one works)\nexport DEEPSEEK_API_KEY=sk-... # recommended: cheapest, $0.002/run\nexport ANTHROPIC_API_KEY=sk-ant-... # Claude\nexport OPENAI_API_KEY=sk-... # GPT-4o\n\n# 3. Run on any project\ncd any-project\ngenesis .\n\n# Or skip approval for CI/automation\ngenesis . --yes\ngenesis . --provider claude\ngenesis . --dry-run # preview without writing\ngenesis . --json # machine-readable output\n```\n\n---\n\n## What Gets Created\n\n```\nyour-project/\n├── prompts/\n│ ├── backend-api-engineer.md # Agent owning backend/ · 36 files\n│ ├── frontend-crafter.md # Agent owning frontend/ · 12 files\n│ ├── db-engineer.md # Agent owning database/ · 8 files\n│ └── reviewer.md # Read-only reviewer · gates all merges\n├── PROJECT-CONTEXT.md # Full AI narrative of your codebase\n└── opencode.json # Registered agent team\n```\n\n---\n\n## How It Works\n\n```\nPhase 1: Survey\nReads directory tree, tech stack, dependencies, imports, configs, tests.\n\nPhase 2: Pre-Analyze (local, 0 tokens, 0 cost)\nDetects hardcoded secrets, AI slop, long files, circular imports,\nmissing error handling, untested files, dead code.\n\nPhase 3: Domain Detection (LLM, ~$0.001)\nSends directory structure + quality flags. LLM identifies natural\ndomain boundaries — the seams in your codebase.\n\nPhase 4: Deep Analysis (LLM, ~$0.005/domain)\nSends EVERY file in each domain. LLM reads every line. Finds bugs at\nfile:line precision. Detects AI-generated code. Flags security issues.\nGenerates the agent prompt with all knowledge extracted from real code.\n\nPhase 5: Build (local, 0 tokens)\nRenders agent .md files, writes PROJECT-CONTEXT.md, registers team\nin opencode.json.\n\nApproval: You review findings before anything is written. Default yes.\n```\n\n---\n\n## Providers\n\n| Provider | Model | Cost/Run* | Context | Setup |\n|----------|-------|-----------|---------|-------|\n| **DeepSeek** | V3 | $0.002 | 128K | `export DEEPSEEK_API_KEY=sk-...` |\n| Claude | Sonnet 4 | $0.08 | 200K | `export ANTHROPIC_API_KEY=sk-ant-...` |\n| OpenAI | GPT-4o | $0.05 | 128K | `export OPENAI_API_KEY=sk-...` |\n\n*For a 120-file project. Genesis auto-detects which provider to use from your environment.\n\n[Full cost breakdown →](docs/TOKEN-COSTS.md)\n\n---\n\n## MCP Server\n\nGenesis also runs as an MCP server — connect it to OpenCode, Claude Code, or Cursor:\n\n```json\n{\n \"mcpServers\": {\n \"genesis\": {\n \"command\": \"npx\",\n \"args\": [\"genesis-mcp\"]\n }\n }\n}\n```\n\nTools: `survey_project`, `propose_team`, `build_team`. Run Genesis from inside your editor — no terminal needed.\n\n---\n\n## Principles\n\n- **Every file is read.** No sampling for quality assessment. The LLM sees your actual code.\n- **No token limits.** Choose DeepSeek and it's effectively free (~500 runs per $1).\n- **Your code stays local.** Sent directly to the API you choose. Never cached. Never stored.\n- **Reviewer always included.** Every team gets a read-only gate. Non-negotiable.\n- **File-scoped permissions.** Agents can only edit their domain. Real filesystem paths.\n\n---\n\n## Docs\n\n| Doc | What |\n|-----|------|\n| [PRD](docs/PRD.md) | What, why, who, success metrics |\n| [Architecture](docs/ARCHITECTURE.md) | System design, data flow, provider model |\n| [Rules](docs/RULES.md) | How Genesis builds itself using its own agents |\n| [Token Costs](docs/TOKEN-COSTS.md) | Per-provider cost table, no-limit strategy |\n| [Roadmap](docs/ROADMAP.md) | 14-day plan, known blockers, mitigations |\n| [Scenarios](docs/SCENARIOS.md) | 5 real-world use cases with terminal output |\n| [Difficulties](docs/DIFFICULTIES.md) | What will break and how we handle it |\n\n---\n\nMIT License · GenesisX v0.2\n"}, "0bd41cca86bf53d20b8c791dabb02ef585743799": {"t": 1785766230.5893207, "n": 9350, "h": "bc5e5910c6200aed", "c": "# GenesisX\n\n> **Every project begins here.** GenesisX reads your entire codebase and builds a tailored team of AI agents — each owning one domain, carrying real architecture knowledge, and scoped with file-level permissions.\n\n> **v0.2** — Enhanced prompt design framework with scoring rubric, anti-patterns, and file priority system. Agent output quality: 8-9/10.\n\n```bash\nnpm install -g genesisx\ncd any-project\ngenesis .\n```\n\n---\n\n## Real Results — 5 GitHub Repositories Tested\n\n### fastapi (Python, 2935 files)\n\n```\n┌──────────────────────────────────────────────┐\n│ fastapi · Python · 8 files │\n│ DeepSeek V3 · ~$0.002 · 5 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ fastapi/ │\n│ tests/ │\n│ docs/ docs_src/ │\n│ scripts/ │\n│ fastapi-slim/ │\n│ 8 bugs · 14 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ fastapi-core-engineer fastapi/ 49 files│\n│ fastapi-test-engineer tests/ 581│\n│ documentation-agent docs/ 2235│\n│ scripts-maintainer scripts/ 69│\n│ fastapi-slim-maintainer fastapi-slim/ 1│\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\nGenerated agent: `fastapi-core-engineer.md` (85 lines). Knows the exact FastAPI version (0.136.3), every core module, all import conventions, async patterns, test framework, and deprecation strategy — extracted straight from the source.\n\n### gin (Go, 53 files) — 2 bugs found\n\n```\n┌──────────────────────────────────────────────┐\n│ gin · Go · 53 files │\n│ DeepSeek V3 · ~$0.002 · 5 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ binding/ render/ internal/ ginS/ │\n│ codec/ docs/ examples/ testdata/ │\n│ 2 bugs · 4 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ gin-core-engineer core modules 54 │\n│ codec-dev codec/ 5 │\n│ docs-agent docs/ 1 │\n│ examples-maintainer examples/ 1 │\n│ testdata-domain-agent testdata/ 7 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n### express (TypeScript, 10 entries) — 1 bug found\n\n```\n┌──────────────────────────────────────────────┐\n│ express · TypeScript (Express) · 10 files │\n│ DeepSeek V3 · ~$0.002 · 3 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ lib/ examples/ test/ │\n│ 1 bug · 8 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ express-core-engineer lib/ 6 │\n│ express-examples-agent examples/ 80 │\n│ express-test-engineer test/ 112 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n### click (Python, 9 entries) — clean codebase\n\n```\n┌──────────────────────────────────────────────┐\n│ click · Python · 9 files │\n│ DeepSeek V3 · ~$0.002 · 4 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ src/ docs/ examples/ tests/ │\n│ 0 bugs · 1 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ core-dev src/ 18 │\n│ click-documentation-agent docs/ 41 │\n│ click-examples-auditor examples/ 39 │\n│ click-test-engineer tests/ 31 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n### got (TypeScript, 9 entries) — 3 bugs + 1 secret\n\n```\n┌──────────────────────────────────────────────┐\n│ got · TypeScript (Express) · 9 files │\n│ DeepSeek V3 · ~$0.002 · 5 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ source/ test/ documentation/ │\n│ media/ benchmark/ │\n│ 3 bugs · 1 secret · 2 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ got-source-agent source/ 23 │\n│ got-test-domain-agent test/ 49 │\n│ documentation-agent documentation/ 26 │\n│ media-asset-manager media/ 4 │\n│ benchmark-agent benchmark/ 2 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n---\n\n## Quick Start\n\n```bash\n# 1. Install\nnpm install -g genesisx\n\n# 2. Set your LLM key (any one works)\nexport DEEPSEEK_API_KEY=sk-... # recommended: cheapest, $0.002/run\nexport ANTHROPIC_API_KEY=sk-ant-... # Claude\nexport OPENAI_API_KEY=sk-... # GPT-4o\n\n# 3. Run on any project\ncd any-project\ngenesis .\n\n# Or skip approval for CI/automation\ngenesis . --yes\ngenesis . --provider claude\ngenesis . --dry-run # preview without writing\ngenesis . --json # machine-readable output\n```\n\n---\n\n## What Gets Created\n\n```\nyour-project/\n├── prompts/\n│ ├── backend-api-engineer.md # Agent owning backend/ · 36 files\n│ ├── frontend-crafter.md # Agent owning frontend/ · 12 files\n│ ├── db-engineer.md # Agent owning database/ · 8 files\n│ └── reviewer.md # Read-only reviewer · gates all merges\n├── PROJECT-CONTEXT.md # Full AI narrative of your codebase\n└── opencode.json # Registered agent team\n```\n\n---\n\n## How It Works\n\n```\nPhase 1: Survey\nReads directory tree, tech stack, dependencies, imports, configs, tests.\n\nPhase 2: Pre-Analyze (local, 0 tokens, 0 cost)\nDetects hardcoded secrets, AI slop, long files, circular imports,\nmissing error handling, untested files, dead code.\n\nPhase 3: Domain Detection (LLM, ~$0.001)\nSends directory structure + quality flags. LLM identifies natural\ndomain boundaries — the seams in your codebase.\n\nPhase 4: Deep Analysis (LLM, ~$0.005/domain)\nSends EVERY file in each domain. LLM reads every line. Finds bugs at\nfile:line precision. Detects AI-generated code. Flags security issues.\nGenerates the agent prompt with all knowledge extracted from real code.\n\nPhase 5: Build (local, 0 tokens)\nRenders agent .md files, writes PROJECT-CONTEXT.md, registers team\nin opencode.json.\n\nApproval: You review findings before anything is written. Default yes.\n```\n\n---\n\n## Providers\n\n| Provider | Model | Cost/Run* | Context | Setup |\n|----------|-------|-----------|---------|-------|\n| **DeepSeek** | V3 | $0.002 | 128K | `export DEEPSEEK_API_KEY=sk-...` |\n| Claude | Sonnet 4 | $0.08 | 200K | `export ANTHROPIC_API_KEY=sk-ant-...` |\n| OpenAI | GPT-4o | $0.05 | 128K | `export OPENAI_API_KEY=sk-...` |\n\n*For a 120-file project. Genesis auto-detects which provider to use from your environment.\n\n[Full cost breakdown →](docs/TOKEN-COSTS.md)\n\n---\n\n## MCP Server\n\nGenesis also runs as an MCP server — connect it to OpenCode, Claude Code, or Cursor:\n\n```json\n{\n \"mcpServers\": {\n \"genesis\": {\n \"command\": \"npx\",\n \"args\": [\"genesis-mcp\"]\n }\n }\n}\n```\n\nTools: `survey_project`, `propose_team`, `build_team`. Run Genesis from inside your editor — no terminal needed.\n\n---\n\n## Principles\n\n- **Every file is read.** No sampling for quality assessment. The LLM sees your actual code.\n- **No token limits.** Choose DeepSeek and it's effectively free (~500 runs per $1).\n- **Your code stays local.** Sent directly to the API you choose. Never cached. Never stored.\n- **Reviewer always included.** Every team gets a read-only gate. Non-negotiable.\n- **File-scoped permissions.** Agents can only edit their domain. Real filesystem paths.\n\n---\n\n## Docs\n\n| Doc | What |\n|-----|------|\n| [PRD](docs/PRD.md) | What, why, who, success metrics |\n| [Architecture](docs/ARCHITECTURE.md) | System design, data flow, provider model |\n| [Rules](docs/RULES.md) | How Genesis builds itself using its own agents |\n| [Token Costs](docs/TOKEN-COSTS.md) | Per-provider cost table, no-limit strategy |\n| [Roadmap](docs/ROADMAP.md) | 14-day plan, known blockers, mitigations |\n| [Scenarios](docs/SCENARIOS.md) | 5 real-world use cases with terminal output |\n| [Difficulties](docs/DIFFICULTIES.md) | What will break and how we handle it |\n\n---\n\nMIT License · GenesisX v0.2\n"}, "9a72832f2e6abac8a1e0a57208505c9d6d7e6610": {"t": 1785766232.085994, "n": 9221, "h": "f2dbca7adda40db0", "c": "# Genesis\n\n> **Every project begins here.** Genesis reads your entire codebase and builds a tailored team of AI agents — each owning one domain, carrying real architecture knowledge, and scoped with file-level permissions.\n\n```bash\nnpm install -g genesisx\ncd any-project\ngenesis .\n```\n\n---\n\n## Real Results — 5 GitHub Repositories Tested\n\n### fastapi (Python, 2935 files)\n\n```\n┌──────────────────────────────────────────────┐\n│ fastapi · Python · 8 files │\n│ DeepSeek V3 · ~$0.002 · 5 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ fastapi/ │\n│ tests/ │\n│ docs/ docs_src/ │\n│ scripts/ │\n│ fastapi-slim/ │\n│ 8 bugs · 14 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ fastapi-core-engineer fastapi/ 49 files│\n│ fastapi-test-engineer tests/ 581│\n│ documentation-agent docs/ 2235│\n│ scripts-maintainer scripts/ 69│\n│ fastapi-slim-maintainer fastapi-slim/ 1│\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\nGenerated agent: `fastapi-core-engineer.md` (85 lines). Knows the exact FastAPI version (0.136.3), every core module, all import conventions, async patterns, test framework, and deprecation strategy — extracted straight from the source.\n\n### gin (Go, 53 files) — 2 bugs found\n\n```\n┌──────────────────────────────────────────────┐\n│ gin · Go · 53 files │\n│ DeepSeek V3 · ~$0.002 · 5 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ binding/ render/ internal/ ginS/ │\n│ codec/ docs/ examples/ testdata/ │\n│ 2 bugs · 4 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ gin-core-engineer core modules 54 │\n│ codec-dev codec/ 5 │\n│ docs-agent docs/ 1 │\n│ examples-maintainer examples/ 1 │\n│ testdata-domain-agent testdata/ 7 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n### express (TypeScript, 10 entries) — 1 bug found\n\n```\n┌──────────────────────────────────────────────┐\n│ express · TypeScript (Express) · 10 files │\n│ DeepSeek V3 · ~$0.002 · 3 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ lib/ examples/ test/ │\n│ 1 bug · 8 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ express-core-engineer lib/ 6 │\n│ express-examples-agent examples/ 80 │\n│ express-test-engineer test/ 112 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n### click (Python, 9 entries) — clean codebase\n\n```\n┌──────────────────────────────────────────────┐\n│ click · Python · 9 files │\n│ DeepSeek V3 · ~$0.002 · 4 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ src/ docs/ examples/ tests/ │\n│ 0 bugs · 1 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ core-dev src/ 18 │\n│ click-documentation-agent docs/ 41 │\n│ click-examples-auditor examples/ 39 │\n│ click-test-engineer tests/ 31 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n### got (TypeScript, 9 entries) — 3 bugs + 1 secret\n\n```\n┌──────────────────────────────────────────────┐\n│ got · TypeScript (Express) · 9 files │\n│ DeepSeek V3 · ~$0.002 · 5 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ source/ test/ documentation/ │\n│ media/ benchmark/ │\n│ 3 bugs · 1 secret · 2 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ got-source-agent source/ 23 │\n│ got-test-domain-agent test/ 49 │\n│ documentation-agent documentation/ 26 │\n│ media-asset-manager media/ 4 │\n│ benchmark-agent benchmark/ 2 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n---\n\n## Quick Start\n\n```bash\n# 1. Install\nnpm install -g genesisx\n\n# 2. Set your LLM key (any one works)\nexport DEEPSEEK_API_KEY=sk-... # recommended: cheapest, $0.002/run\nexport ANTHROPIC_API_KEY=sk-ant-... # Claude\nexport OPENAI_API_KEY=sk-... # GPT-4o\n\n# 3. Run on any project\ncd any-project\ngenesis .\n\n# Or skip approval for CI/automation\ngenesis . --yes\ngenesis . --provider claude\ngenesis . --dry-run # preview without writing\ngenesis . --json # machine-readable output\n```\n\n---\n\n## What Gets Created\n\n```\nyour-project/\n├── prompts/\n│ ├── backend-api-engineer.md # Agent owning backend/ · 36 files\n│ ├── frontend-crafter.md # Agent owning frontend/ · 12 files\n│ ├── db-engineer.md # Agent owning database/ · 8 files\n│ └── reviewer.md # Read-only reviewer · gates all merges\n├── PROJECT-CONTEXT.md # Full AI narrative of your codebase\n└── opencode.json # Registered agent team\n```\n\n---\n\n## How It Works\n\n```\nPhase 1: Survey\nReads directory tree, tech stack, dependencies, imports, configs, tests.\n\nPhase 2: Pre-Analyze (local, 0 tokens, 0 cost)\nDetects hardcoded secrets, AI slop, long files, circular imports,\nmissing error handling, untested files, dead code.\n\nPhase 3: Domain Detection (LLM, ~$0.001)\nSends directory structure + quality flags. LLM identifies natural\ndomain boundaries — the seams in your codebase.\n\nPhase 4: Deep Analysis (LLM, ~$0.005/domain)\nSends EVERY file in each domain. LLM reads every line. Finds bugs at\nfile:line precision. Detects AI-generated code. Flags security issues.\nGenerates the agent prompt with all knowledge extracted from real code.\n\nPhase 5: Build (local, 0 tokens)\nRenders agent .md files, writes PROJECT-CONTEXT.md, registers team\nin opencode.json.\n\nApproval: You review findings before anything is written. Default yes.\n```\n\n---\n\n## Providers\n\n| Provider | Model | Cost/Run* | Context | Setup |\n|----------|-------|-----------|---------|-------|\n| **DeepSeek** | V3 | $0.002 | 128K | `export DEEPSEEK_API_KEY=sk-...` |\n| Claude | Sonnet 4 | $0.08 | 200K | `export ANTHROPIC_API_KEY=sk-ant-...` |\n| OpenAI | GPT-4o | $0.05 | 128K | `export OPENAI_API_KEY=sk-...` |\n\n*For a 120-file project. Genesis auto-detects which provider to use from your environment.\n\n[Full cost breakdown →](docs/TOKEN-COSTS.md)\n\n---\n\n## MCP Server\n\nGenesis also runs as an MCP server — connect it to OpenCode, Claude Code, or Cursor:\n\n```json\n{\n \"mcpServers\": {\n \"genesis\": {\n \"command\": \"npx\",\n \"args\": [\"genesis-mcp\"]\n }\n }\n}\n```\n\nTools: `survey_project`, `propose_team`, `build_team`. Run Genesis from inside your editor — no terminal needed.\n\n---\n\n## Principles\n\n- **Every file is read.** No sampling for quality assessment. The LLM sees your actual code.\n- **No token limits.** Choose DeepSeek and it's effectively free (~500 runs per $1).\n- **Your code stays local.** Sent directly to the API you choose. Never cached. Never stored.\n- **Reviewer always included.** Every team gets a read-only gate. Non-negotiable.\n- **File-scoped permissions.** Agents can only edit their domain. Real filesystem paths.\n\n---\n\n## Docs\n\n| Doc | What |\n|-----|------|\n| [PRD](docs/PRD.md) | What, why, who, success metrics |\n| [Architecture](docs/ARCHITECTURE.md) | System design, data flow, provider model |\n| [Rules](docs/RULES.md) | How Genesis builds itself using its own agents |\n| [Token Costs](docs/TOKEN-COSTS.md) | Per-provider cost table, no-limit strategy |\n| [Roadmap](docs/ROADMAP.md) | 14-day plan, known blockers, mitigations |\n| [Scenarios](docs/SCENARIOS.md) | 5 real-world use cases with terminal output |\n| [Difficulties](docs/DIFFICULTIES.md) | What will break and how we handle it |\n\n---\n\nMIT License · Built with Genesis itself\n"}, "f1491890bde615f458a1536791c662672fcb191c": {"t": 1785766233.4722054, "n": 9222, "h": "665cd9d0c48320fc", "c": "# Genesis\n\n> **Every project begins here.** Genesis reads your entire codebase and builds a tailored team of AI agents — each owning one domain, carrying real architecture knowledge, and scoped with file-level permissions.\n\n```bash\nnpm install -g genesisx\ncd any-project\ngenesis .\n```\n\n---\n\n## Real Results — 5 GitHub Repositories Tested\n\n### fastapi (Python, 2935 files)\n\n```\n┌──────────────────────────────────────────────┐\n│ fastapi · Python · 8 files │\n│ DeepSeek V3 · ~$0.002 · 5 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ fastapi/ │\n│ tests/ │\n│ docs/ docs_src/ │\n│ scripts/ │\n│ fastapi-slim/ │\n│ 8 bugs · 14 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ fastapi-core-engineer fastapi/ 49 files│\n│ fastapi-test-engineer tests/ 581│\n│ documentation-agent docs/ 2235│\n│ scripts-maintainer scripts/ 69│\n│ fastapi-slim-maintainer fastapi-slim/ 1│\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\nGenerated agent: `fastapi-core-engineer.md` (85 lines). Knows the exact FastAPI version (0.136.3), every core module, all import conventions, async patterns, test framework, and deprecation strategy — extracted straight from the source.\n\n### gin (Go, 53 files) — 2 bugs found\n\n```\n┌──────────────────────────────────────────────┐\n│ gin · Go · 53 files │\n│ DeepSeek V3 · ~$0.002 · 5 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ binding/ render/ internal/ ginS/ │\n│ codec/ docs/ examples/ testdata/ │\n│ 2 bugs · 4 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ gin-core-engineer core modules 54 │\n│ codec-dev codec/ 5 │\n│ docs-agent docs/ 1 │\n│ examples-maintainer examples/ 1 │\n│ testdata-domain-agent testdata/ 7 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n### express (TypeScript, 10 entries) — 1 bug found\n\n```\n┌──────────────────────────────────────────────┐\n│ express · TypeScript (Express) · 10 files │\n│ DeepSeek V3 · ~$0.002 · 3 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ lib/ examples/ test/ │\n│ 1 bug · 8 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ express-core-engineer lib/ 6 │\n│ express-examples-agent examples/ 80 │\n│ express-test-engineer test/ 112 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n### click (Python, 9 entries) — clean codebase\n\n```\n┌──────────────────────────────────────────────┐\n│ click · Python · 9 files │\n│ DeepSeek V3 · ~$0.002 · 4 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ src/ docs/ examples/ tests/ │\n│ 0 bugs · 1 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ core-dev src/ 18 │\n│ click-documentation-agent docs/ 41 │\n│ click-examples-auditor examples/ 39 │\n│ click-test-engineer tests/ 31 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n### got (TypeScript, 9 entries) — 3 bugs + 1 secret\n\n```\n┌──────────────────────────────────────────────┐\n│ got · TypeScript (Express) · 9 files │\n│ DeepSeek V3 · ~$0.002 · 5 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ source/ test/ documentation/ │\n│ media/ benchmark/ │\n│ 3 bugs · 1 secret · 2 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ got-source-agent source/ 23 │\n│ got-test-domain-agent test/ 49 │\n│ documentation-agent documentation/ 26 │\n│ media-asset-manager media/ 4 │\n│ benchmark-agent benchmark/ 2 │\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\n---\n\n## Quick Start\n\n```bash\n# 1. Install\nnpm install -g genesisx\n\n# 2. Set your LLM key (any one works)\nexport DEEPSEEK_API_KEY=sk-... # recommended: cheapest, $0.002/run\nexport ANTHROPIC_API_KEY=sk-ant-... # Claude\nexport OPENAI_API_KEY=sk-... # GPT-4o\n\n# 3. Run on any project\ncd any-project\ngenesis .\n\n# Or skip approval for CI/automation\ngenesis . --yes\ngenesis . --provider claude\ngenesis . --dry-run # preview without writing\ngenesis . --json # machine-readable output\n```\n\n---\n\n## What Gets Created\n\n```\nyour-project/\n├── prompts/\n│ ├── backend-api-engineer.md # Agent owning backend/ · 36 files\n│ ├── frontend-crafter.md # Agent owning frontend/ · 12 files\n│ ├── db-engineer.md # Agent owning database/ · 8 files\n│ └── reviewer.md # Read-only reviewer · gates all merges\n├── PROJECT-CONTEXT.md # Full AI narrative of your codebase\n└── opencode.json # Registered agent team\n```\n\n---\n\n## How It Works\n\n```\nPhase 1: Survey\nReads directory tree, tech stack, dependencies, imports, configs, tests.\n\nPhase 2: Pre-Analyze (local, 0 tokens, 0 cost)\nDetects hardcoded secrets, AI slop, long files, circular imports,\nmissing error handling, untested files, dead code.\n\nPhase 3: Domain Detection (LLM, ~$0.001)\nSends directory structure + quality flags. LLM identifies natural\ndomain boundaries — the seams in your codebase.\n\nPhase 4: Deep Analysis (LLM, ~$0.005/domain)\nSends EVERY file in each domain. LLM reads every line. Finds bugs at\nfile:line precision. Detects AI-generated code. Flags security issues.\nGenerates the agent prompt with all knowledge extracted from real code.\n\nPhase 5: Build (local, 0 tokens)\nRenders agent .md files, writes PROJECT-CONTEXT.md, registers team\nin opencode.json.\n\nApproval: You review findings before anything is written. Default yes.\n```\n\n---\n\n## Providers\n\n| Provider | Model | Cost/Run* | Context | Setup |\n|----------|-------|-----------|---------|-------|\n| **DeepSeek** | V3 | $0.002 | 128K | `export DEEPSEEK_API_KEY=sk-...` |\n| Claude | Sonnet 4 | $0.08 | 200K | `export ANTHROPIC_API_KEY=sk-ant-...` |\n| OpenAI | GPT-4o | $0.05 | 128K | `export OPENAI_API_KEY=sk-...` |\n\n*For a 120-file project. Genesis auto-detects which provider to use from your environment.\n\n[Full cost breakdown →](docs/TOKEN-COSTS.md)\n\n---\n\n## MCP Server\n\nGenesis also runs as an MCP server — connect it to OpenCode, Claude Code, or Cursor:\n\n```json\n{\n \"mcpServers\": {\n \"genesis\": {\n \"command\": \"npx\",\n \"args\": [\"genesisx-mcp\"]\n }\n }\n}\n```\n\nTools: `survey_project`, `propose_team`, `build_team`. Run Genesis from inside your editor — no terminal needed.\n\n---\n\n## Principles\n\n- **Every file is read.** No sampling for quality assessment. The LLM sees your actual code.\n- **No token limits.** Choose DeepSeek and it's effectively free (~500 runs per $1).\n- **Your code stays local.** Sent directly to the API you choose. Never cached. Never stored.\n- **Reviewer always included.** Every team gets a read-only gate. Non-negotiable.\n- **File-scoped permissions.** Agents can only edit their domain. Real filesystem paths.\n\n---\n\n## Docs\n\n| Doc | What |\n|-----|------|\n| [PRD](docs/PRD.md) | What, why, who, success metrics |\n| [Architecture](docs/ARCHITECTURE.md) | System design, data flow, provider model |\n| [Rules](docs/RULES.md) | How Genesis builds itself using its own agents |\n| [Token Costs](docs/TOKEN-COSTS.md) | Per-provider cost table, no-limit strategy |\n| [Roadmap](docs/ROADMAP.md) | 14-day plan, known blockers, mitigations |\n| [Scenarios](docs/SCENARIOS.md) | 5 real-world use cases with terminal output |\n| [Difficulties](docs/DIFFICULTIES.md) | What will break and how we handle it |\n\n---\n\nMIT License · Built with Genesis itself\n"}, "bafd7b7152861864021e4f72b4a62e3967d7019a": {"t": 1785766234.6521125, "n": 6196, "h": "5295b3627480fea2", "c": "# Genesis\n\n> **Every project begins here.** Genesis reads your entire codebase and builds a tailored team of AI agents — each owning one domain, carrying real architecture knowledge, and scoped with file-level permissions.\n\n```bash\nnpm install -g agent-genesis\ncd any-project\ngenesis .\n```\n\n---\n\n## Real Results — 5 GitHub Repositories Tested\n\n### fastapi (Python, 2935 files)\n\n```\n┌──────────────────────────────────────────────┐\n│ fastapi · Python · 8 files │\n│ DeepSeek V3 · ~$0.002 · 5 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ fastapi/ │\n│ tests/ │\n│ docs/ docs_src/ │\n│ scripts/ │\n│ fastapi-slim/ │\n│ 8 bugs · 14 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ fastapi-core-engineer fastapi/ 49 files│\n│ fastapi-test-engineer tests/ 581│\n│ documentation-agent docs/ 2235│\n│ scripts-maintainer scripts/ 69│\n│ fastapi-slim-maintainer fastapi-slim/ 1│\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\nGenerated agent: `fastapi-core-engineer.md` (85 lines). Knows the exact FastAPI version (0.136.3), every core module, all import conventions, async patterns, test framework, and deprecation strategy — extracted straight from the source.\n\n### gin (Go, 53 files)\n\nFound 2 bugs. Generated 5 agents: `gin-core-engineer`, `codec-dev`, `docs-agent`, `examples-maintainer`, `testdata-domain-agent`.\n\n### express (TypeScript / JavaScript, 10 entries)\n\nFound 1 bug. Generated 3 agents: `express-core-engineer`, `express-examples-agent`, `express-test-engineer`.\n\n### click (Python, 9 entries)\n\nZero bugs (clean codebase). 4 agents: `core-dev`, `click-documentation-agent`, `click-examples-auditor`, `click-test-engineer`.\n\n### got (TypeScript, 9 entries)\n\nFound 3 bugs and 1 hardcoded secret. 5 agents: `got-source-agent`, `got-test-domain-agent`, `documentation-agent`, `media-asset-manager`, `benchmark-agent`.\n\n---\n\n## Quick Start\n\n```bash\n# 1. Install\nnpm install -g genesis\n\n# 2. Set your LLM key (any one works)\nexport DEEPSEEK_API_KEY=sk-... # recommended: cheapest, $0.002/run\nexport ANTHROPIC_API_KEY=sk-ant-... # Claude\nexport OPENAI_API_KEY=sk-... # GPT-4o\n\n# 3. Run on any project\ncd any-project\nnpx genesis .\n\n# Or skip approval for CI/automation\nnpx genesis . --yes\nnpx genesis . --provider claude\nnpx genesis . --dry-run # preview without writing\nnpx genesis . --json # machine-readable output\n```\n\n---\n\n## What Gets Created\n\n```\nyour-project/\n├── prompts/\n│ ├── backend-api-engineer.md # Agent owning backend/ · 36 files\n│ ├── frontend-crafter.md # Agent owning frontend/ · 12 files\n│ ├── db-engineer.md # Agent owning database/ · 8 files\n│ └── reviewer.md # Read-only reviewer · gates all merges\n├── PROJECT-CONTEXT.md # Full AI narrative of your codebase\n└── opencode.json # Registered agent team\n```\n\n---\n\n## How It Works\n\n```\nPhase 1: Survey\nReads directory tree, tech stack, dependencies, imports, configs, tests.\n\nPhase 2: Pre-Analyze (local, 0 tokens, 0 cost)\nDetects hardcoded secrets, AI slop, long files, circular imports,\nmissing error handling, untested files, dead code.\n\nPhase 3: Domain Detection (LLM, ~$0.001)\nSends directory structure + quality flags. LLM identifies natural\ndomain boundaries — the seams in your codebase.\n\nPhase 4: Deep Analysis (LLM, ~$0.005/domain)\nSends EVERY file in each domain. LLM reads every line. Finds bugs at\nfile:line precision. Detects AI-generated code. Flags security issues.\nGenerates the agent prompt with all knowledge extracted from real code.\n\nPhase 5: Build (local, 0 tokens)\nRenders agent .md files, writes PROJECT-CONTEXT.md, registers team\nin opencode.json.\n\nApproval: You review findings before anything is written. Default yes.\n```\n\n---\n\n## Providers\n\n| Provider | Model | Cost/Run* | Context | Setup |\n|----------|-------|-----------|---------|-------|\n| **DeepSeek** | V3 | $0.002 | 128K | `export DEEPSEEK_API_KEY=sk-...` |\n| Claude | Sonnet 4 | $0.08 | 200K | `export ANTHROPIC_API_KEY=sk-ant-...` |\n| OpenAI | GPT-4o | $0.05 | 128K | `export OPENAI_API_KEY=sk-...` |\n\n*For a 120-file project. Genesis auto-detects which provider to use from your environment.\n\n[Full cost breakdown →](docs/TOKEN-COSTS.md)\n\n---\n\n## MCP Server\n\nGenesis also runs as an MCP server — connect it to OpenCode, Claude Code, or Cursor:\n\n```json\n{\n \"mcpServers\": {\n \"genesis\": {\n \"command\": \"npx\",\n \"args\": [\"genesis-mcp\"]\n }\n }\n}\n```\n\nTools: `survey_project`, `propose_team`, `build_team`. Run Genesis from inside your editor — no terminal needed.\n\n---\n\n## Principles\n\n- **Every file is read.** No sampling for quality assessment. The LLM sees your actual code.\n- **No token limits.** Choose DeepSeek and it's effectively free (~500 runs per $1).\n- **Your code stays local.** Sent directly to the API you choose. Never cached. Never stored.\n- **Reviewer always included.** Every team gets a read-only gate. Non-negotiable.\n- **File-scoped permissions.** Agents can only edit their domain. Real filesystem paths.\n\n---\n\n## Docs\n\n| Doc | What |\n|-----|------|\n| [PRD](docs/PRD.md) | What, why, who, success metrics |\n| [Architecture](docs/ARCHITECTURE.md) | System design, data flow, provider model |\n| [Rules](docs/RULES.md) | How Genesis builds itself using its own agents |\n| [Token Costs](docs/TOKEN-COSTS.md) | Per-provider cost table, no-limit strategy |\n| [Roadmap](docs/ROADMAP.md) | 14-day plan, known blockers, mitigations |\n| [Scenarios](docs/SCENARIOS.md) | 5 real-world use cases with terminal output |\n| [Difficulties](docs/DIFFICULTIES.md) | What will break and how we handle it |\n\n---\n\nMIT License · Built with Genesis itself\n"}, "fdd02ffc15a9ec0d332542467384b5b3ee4b6705": {"t": 1785766236.0062764, "n": 6194, "h": "0fbe5f30060a6b85", "c": "# Genesis\n\n> **Every project begins here.** Genesis reads your entire codebase and builds a tailored team of AI agents — each owning one domain, carrying real architecture knowledge, and scoped with file-level permissions.\n\n```bash\nnpm install -g genesis\ncd any-project\nnpx genesis .\n```\n\n---\n\n## Real Results — 5 GitHub Repositories Tested\n\n### fastapi (Python, 2935 files)\n\n```\n┌──────────────────────────────────────────────┐\n│ fastapi · Python · 8 files │\n│ DeepSeek V3 · ~$0.002 · 5 domains │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ files │\n│ fastapi/ │\n│ tests/ │\n│ docs/ docs_src/ │\n│ scripts/ │\n│ fastapi-slim/ │\n│ 8 bugs · 14 untested │\n└──────────────────────────────────────────────┘\n\n┌──────────────────────────────────────────────┐\n│ team │\n│ fastapi-core-engineer fastapi/ 49 files│\n│ fastapi-test-engineer tests/ 581│\n│ documentation-agent docs/ 2235│\n│ scripts-maintainer scripts/ 69│\n│ fastapi-slim-maintainer fastapi-slim/ 1│\n│ reviewer all · read-only │\n└──────────────────────────────────────────────┘\n```\n\nGenerated agent: `fastapi-core-engineer.md` (85 lines). Knows the exact FastAPI version (0.136.3), every core module, all import conventions, async patterns, test framework, and deprecation strategy — extracted straight from the source.\n\n### gin (Go, 53 files)\n\nFound 2 bugs. Generated 5 agents: `gin-core-engineer`, `codec-dev`, `docs-agent`, `examples-maintainer`, `testdata-domain-agent`.\n\n### express (TypeScript / JavaScript, 10 entries)\n\nFound 1 bug. Generated 3 agents: `express-core-engineer`, `express-examples-agent`, `express-test-engineer`.\n\n### click (Python, 9 entries)\n\nZero bugs (clean codebase). 4 agents: `core-dev`, `click-documentation-agent`, `click-examples-auditor`, `click-test-engineer`.\n\n### got (TypeScript, 9 entries)\n\nFound 3 bugs and 1 hardcoded secret. 5 agents: `got-source-agent`, `got-test-domain-agent`, `documentation-agent`, `media-asset-manager`, `benchmark-agent`.\n\n---\n\n## Quick Start\n\n```bash\n# 1. Install\nnpm install -g genesis\n\n# 2. Set your LLM key (any one works)\nexport DEEPSEEK_API_KEY=sk-... # recommended: cheapest, $0.002/run\nexport ANTHROPIC_API_KEY=sk-ant-... # Claude\nexport OPENAI_API_KEY=sk-... # GPT-4o\n\n# 3. Run on any project\ncd any-project\nnpx genesis .\n\n# Or skip approval for CI/automation\nnpx genesis . --yes\nnpx genesis . --provider claude\nnpx genesis . --dry-run # preview without writing\nnpx genesis . --json # machine-readable output\n```\n\n---\n\n## What Gets Created\n\n```\nyour-project/\n├── prompts/\n│ ├── backend-api-engineer.md # Agent owning backend/ · 36 files\n│ ├── frontend-crafter.md # Agent owning frontend/ · 12 files\n│ ├── db-engineer.md # Agent owning database/ · 8 files\n│ └── reviewer.md # Read-only reviewer · gates all merges\n├── PROJECT-CONTEXT.md # Full AI narrative of your codebase\n└── opencode.json # Registered agent team\n```\n\n---\n\n## How It Works\n\n```\nPhase 1: Survey\nReads directory tree, tech stack, dependencies, imports, configs, tests.\n\nPhase 2: Pre-Analyze (local, 0 tokens, 0 cost)\nDetects hardcoded secrets, AI slop, long files, circular imports,\nmissing error handling, untested files, dead code.\n\nPhase 3: Domain Detection (LLM, ~$0.001)\nSends directory structure + quality flags. LLM identifies natural\ndomain boundaries — the seams in your codebase.\n\nPhase 4: Deep Analysis (LLM, ~$0.005/domain)\nSends EVERY file in each domain. LLM reads every line. Finds bugs at\nfile:line precision. Detects AI-generated code. Flags security issues.\nGenerates the agent prompt with all knowledge extracted from real code.\n\nPhase 5: Build (local, 0 tokens)\nRenders agent .md files, writes PROJECT-CONTEXT.md, registers team\nin opencode.json.\n\nApproval: You review findings before anything is written. Default yes.\n```\n\n---\n\n## Providers\n\n| Provider | Model | Cost/Run* | Context | Setup |\n|----------|-------|-----------|---------|-------|\n| **DeepSeek** | V3 | $0.002 | 128K | `export DEEPSEEK_API_KEY=sk-...` |\n| Claude | Sonnet 4 | $0.08 | 200K | `export ANTHROPIC_API_KEY=sk-ant-...` |\n| OpenAI | GPT-4o | $0.05 | 128K | `export OPENAI_API_KEY=sk-...` |\n\n*For a 120-file project. Genesis auto-detects which provider to use from your environment.\n\n[Full cost breakdown →](docs/TOKEN-COSTS.md)\n\n---\n\n## MCP Server\n\nGenesis also runs as an MCP server — connect it to OpenCode, Claude Code, or Cursor:\n\n```json\n{\n \"mcpServers\": {\n \"genesis\": {\n \"command\": \"npx\",\n \"args\": [\"genesis-mcp\"]\n }\n }\n}\n```\n\nTools: `survey_project`, `propose_team`, `build_team`. Run Genesis from inside your editor — no terminal needed.\n\n---\n\n## Principles\n\n- **Every file is read.** No sampling for quality assessment. The LLM sees your actual code.\n- **No token limits.** Choose DeepSeek and it's effectively free (~500 runs per $1).\n- **Your code stays local.** Sent directly to the API you choose. Never cached. Never stored.\n- **Reviewer always included.** Every team gets a read-only gate. Non-negotiable.\n- **File-scoped permissions.** Agents can only edit their domain. Real filesystem paths.\n\n---\n\n## Docs\n\n| Doc | What |\n|-----|------|\n| [PRD](docs/PRD.md) | What, why, who, success metrics |\n| [Architecture](docs/ARCHITECTURE.md) | System design, data flow, provider model |\n| [Rules](docs/RULES.md) | How Genesis builds itself using its own agents |\n| [Token Costs](docs/TOKEN-COSTS.md) | Per-provider cost table, no-limit strategy |\n| [Roadmap](docs/ROADMAP.md) | 14-day plan, known blockers, mitigations |\n| [Scenarios](docs/SCENARIOS.md) | 5 real-world use cases with terminal output |\n| [Difficulties](docs/DIFFICULTIES.md) | What will break and how we handle it |\n\n---\n\nMIT License · Built with Genesis itself\n"}, "8c2426c47ecfee42cfcc1ac790cb28ea7ccbfe52": {"t": 1785766237.5750647, "n": 4201, "h": "5ac171189fb3ed2c", "c": "# Genesis\n\n> Every project begins here. Genesis reads your codebase and builds a tailored AI engineering team.\n\n```bash\nnpx genesis .\n```\n\n---\n\n## What It Does\n\nGenesis reads EVERY file in your project. It finds bugs, messy code, AI slop, and security issues. Then it generates a team of AI agents — each owning one domain of your codebase, with file-level permissions and deep architecture knowledge extracted from your actual code.\n\n**One command. Your AI engineering team, built.**\n\n---\n\n## Quick Start\n\n```bash\nnpm install -g genesis\ncd any-project\nnpx genesis .\n```\n\nGenesis will:\n1. Survey your project (language, framework, modules, tests, configs)\n2. Pre-analyze code quality (bugs, secrets, AI slop, complexity)\n3. Call an LLM to identify domain boundaries and deeply analyze every file\n4. Propose an agent team with reasoning\n5. Build the team: `./prompts/*.md`, `PROJECT-CONTEXT.md`, `opencode.json`\n\n---\n\n## Providers\n\nGenesis auto-detects your LLM from environment variables:\n\n```bash\n# DeepSeek (recommended — cheapest, highest quality for analysis)\nexport DEEPSEEK_API_KEY=sk-...\n\n# Claude\nexport ANTHROPIC_API_KEY=sk-ant-...\n\n# OpenAI\nexport OPENAI_API_KEY=sk-...\n```\n\nOr specify explicitly:\n\n```bash\nnpx genesis . --provider claude\nnpx genesis . --provider openai\n```\n\n[Full token cost breakdown →](docs/TOKEN-COSTS.md)\n\n---\n\n## What Gets Created\n\n```\nyour-project/\n├── prompts/\n│ ├── backend-dev.md # Agent owning backend/ domain\n│ ├── frontend-dev.md # Agent owning frontend/ domain\n│ ├── db-engineer.md # Agent owning database/ domain\n│ └── reviewer.md # Read-only reviewer gate\n├── PROJECT-CONTEXT.md # Full AI narrative of your project\n└── opencode.json # Registered agent team\n```\n\n---\n\n## Commands\n\n```bash\nnpx genesis . # Full pipeline\nnpx genesis survey . # Survey only\nnpx genesis . --yes # Skip approval\nnpx genesis . --dry-run # Show what would be created\nnpx genesis . --json # Output as JSON\n```\n\n---\n\n## MCP Server\n\nGenesis also works as an MCP server. Register it in your MCP config:\n\n```json\n{\n \"mcpServers\": {\n \"genesis\": {\n \"command\": \"npx\",\n \"args\": [\"genesis-mcp\"]\n }\n }\n}\n```\n\nAvailable tools: `survey_project`, `propose_team`, `build_team`.\n\n---\n\n## How It Works (Deep)\n\n### Phase A: Structure Survey\nReads directory tree, tech stack, import graph, dependencies, config files. Pre-analyzer runs locally (0 tokens, 0 cost) to detect bugs, secrets, AI slop, complexity.\n\n### Phase A LLM: Domain Detection\nSends structure only (no file contents) to determine natural domain boundaries. ~$0.001.\n\n### Phase B: Per-Domain Deep Analysis\nFor EACH domain, sends ALL files to LLM. Reads every line. Finds bugs, data flow issues, AI slop, security problems. Generates agent .md with architecture knowledge. ~$0.005 per domain.\n\n### Phase C: Build (Local)\nRenders templates, writes agent files, registers team in opencode.json. 0 tokens. 0 cost.\n\n[Full architecture →](docs/ARCHITECTURE.md)\n\n---\n\n## Cost\n\nAt DeepSeek pricing, analyzing a 120-file project costs **$0.02**. A 500-file project costs **$0.04**. Free-tier viable.\n\n[Complete cost table →](docs/TOKEN-COSTS.md)\n\n---\n\n## Engineering Principles\n\n- **Every file is read by the LLM.** No sampling. No shortcuts.\n- **No token limits.** Quality first, always.\n- **Your code never leaves your machine except to the API you choose.**\n- **Guaranteed reviewer.** Every team gets a reviewer gate.\n- **File-scoped permissions.** Agents can only edit their own domain.\n- **Deterministic.** Same codebase generates the same team.\n\n---\n\n## Docs\n\n- [PRD](docs/PRD.md) — What, why, who, success metrics\n- [Architecture](docs/ARCHITECTURE.md) — System design, data flow, decisions\n- [Rules](docs/RULES.md) — How Genesis builds itself\n- [Token Costs](docs/TOKEN-COSTS.md) — Per-provider cost breakdown\n- [Roadmap](docs/ROADMAP.md) — 14-day plan, known blockers\n- [Scenarios](docs/SCENARIOS.md) — 5 real-world use cases\n- [Difficulties](docs/DIFFICULTIES.md) — What will go wrong and how we handle it\n\n---\n\n## Status\n\n**v0.1.0** — Core engine + CLI shipped. MCP server in progress.\n\n---\n\nMIT License\n"}}}} |