[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-82002":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":8,"htmlUrl":8,"language":9,"languages":8,"totalLinesOfCode":8,"stars":10,"forks":11,"watchers":12,"openIssues":13,"contributorsCount":13,"subscribersCount":13,"size":13,"stars1d":13,"stars7d":14,"stars30d":14,"stars90d":13,"forks30d":13,"starsTrendScore":13,"compositeScore":15,"rankGlobal":8,"rankLanguage":8,"license":8,"archived":16,"fork":16,"defaultBranch":17,"hasWiki":16,"hasPages":16,"topics":18,"createdAt":8,"pushedAt":8,"updatedAt":19,"readmeContent":20,"aiSummary":21,"trendingCount":13,"starSnapshotCount":13,"syncStatus":22,"lastSyncTime":23,"discoverSource":24},82002,"opencode-config","grojeda\u002Fopencode-config","grojeda",null,"TypeScript",23,3,21,0,1,1.81,false,"main",[],"2026-06-12 02:04:22","# OpenCode Configuration\n\n[![OpenCode](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FOpenCode-Configuration-blue)](https:\u002F\u002Fopencode.ai)\n[![Plugin](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPlugin-warp--dev-green)](https:\u002F\u002Fgithub.com\u002Fwarp-dot-dev\u002Fopencode-warp)\n[![Agents](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FAgents-9-orange)]()\n[![Skills](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FSkills-10-purple)]()\n[![Commands](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FCommands-4-yellow)]()\n\n---\n\n## Table of Contents\n\n- [Overview](#overview)\n- [Installation & Setup](#installation--setup)\n- [Agents Reference](#agents-reference)\n- [Commands Guide](#commands-guide)\n- [Configuration Details](#configuration-details)\n- [Usage Examples](#usage-examples)\n\n---\n\n## Overview\n\nThis configuration transforms OpenCode into a **multi-agent development system** with:\n\n- **9 specialized agents** organized around a research → plan → review → implement → verify pipeline\n- **10 skills** covering browser automation, frontend best practices (Angular, React, NestJS), testing (Jest, Karma, Playwright), PR review, and token-efficient communication\n- **4 lane-based commands** (`\u002Fquick`, `\u002Fmedium`, `\u002Ffull`, `\u002Fdebug`) that route work through the orchestrator with appropriate rigor\n- **Token compression** via the `caveman` skill for efficient agent-to-agent communication\n- **Plugin integration** with `@warp-dot-dev\u002Fopencode-warp`\n\n### Architecture\n\n``` md\nUser Request\n    │\n    ▼\n┌─────────────────────┐\n│  orchestrator-agent │  ← Routes work, manages lanes, approval gates\n└────────┬────────────┘\n         │\n         │\n         ▼                               \n┌──────────┐  ┌──────────────┐  ┌──────────────────┐\n│ research │  │ planning     │→ │ reviewer         │\n│  agent   │→ │   agent      │← │    agent         │\n└──────────┘  └──────┬───────┘  └──────────────────┘\n                     │\n                     ▼\n              ┌──────────────┐   ┌──────────────────┐\n              │implementation│ → │ verifier-agent   │\n              │   agent      │ ← │   (audit)        │\n              └──────┬───────┘   └──────────────────┘\n                     │\n                     ▼\n              ┌──────────────┐\n              │ test-fixer   │ ← When tests fail\n              │   agent      │\n              └──────────────┘\n```\n\n---\n\n## Prerequisites\n\nBefore installing this OpenCode configuration, make sure you have these tools installed.\n\n### Required\n\n| Tool | Purpose |\n|------|---------|\n| OpenCode | Main CLI\u002Fruntime used by this configuration |\n| Node.js | Required for plugins and skills |\n| npm | Installs package dependencies |\n| Git | Clones and updates this repository |\n\n### Recommended\n\n| Tool | Purpose |\n|------|---------|\n| GitHub CLI (`gh`) | Required for PR review workflows |\n| Warp Terminal | Recommended for `@warp-dot-dev\u002Fopencode-warp` |\n| ripgrep (`rg`) | Faster codebase search |\n| Bash\u002FZsh\u002FPowerShell | Working shell environment |\n\n### Verify Your Environment\n\n```bash\nopencode --version\nnode --version\nnpm --version\ngit --version\n```\n---\n\n## Installation & Setup\n\n### Using This Configuration\n\nThis configuration lives in `~\u002F.config\u002Fopencode`. To apply it:\n\n1. **Backup your existing config** (if any):\n\n   ```bash\n   cp -r ~\u002F.config\u002Fopencode ~\u002F.config\u002Fopencode.backup\n   ```\n\n2. **Clone or copy this configuration**:\n\n   ```bash\n   git clone \u003Crepository-url> ~\u002F.config\u002Fopencode-new\n   cp -r ~\u002F.config\u002Fopencode-new\u002F* ~\u002F.config\u002Fopencode\u002F\n   ```\n\n3. **Verify**:\n\n   ```bash\n   opencode --version\n   ```\n\n### Configuration Structure\n\n``` md\n\n~\u002F.config\u002Fopencode\u002F\n├── opencode.json              # Main config: agents, permissions, plugin\n├── tui.json                   # TUI settings\n├── agents\u002F                    # Agent definitions (9 agents)\n│   ├── orchestrator-agent.md\n│   ├── research-agent.md\n│   ├── planning-agent.md\n│   ├── reviewer-agent.md\n│   ├── implementation-agent.md\n│   ├── test-fixer-agent.md\n│   ├── verifier-agent.md\n│   ├── pr-review-agent.md\n│   └── prompt-agent.md\n├── commands\u002F                  # Lane commands (4 commands)\n│   ├── quick.md\n│   ├── medium.md\n│   ├── full.md\n│   └── debug.md\n├── skills\u002F                    # Installed skills (10 skills)\n│   ├── agent-browser\u002F\n│   ├── angular-best-practices\u002F\n│   ├── caveman\u002F\n│   ├── find-skills\u002F\n│   ├── jest-testing\u002F\n│   ├── karma-testing\u002F\n│   ├── nestjs-best-practices\u002F\n│   ├── playwright-cli\u002F\n│   ├── pr-review\u002F\n│   └── vercel-react-best-practices\u002F\n└── package.json               # Skill dependencies\n```\n\n---\n\n## Agent Authoring Workflow\n\nAgent prompts are authored in TypeScript and rendered to Markdown for OpenCode.\n\n| Path | Role |\n|------|------|\n| `src\u002Fagents\u002F` | Source of truth for agent prompts, sections, frontmatter, and output contracts |\n| `src\u002Fconfig\u002F` | Source of truth for agent metadata, models, and routing |\n| `src\u002Fadapters\u002Fmarkdown-agent.ts` | Renderer from typed agent definitions to OpenCode Markdown |\n| `agents\u002F*.md` | Generated runtime files consumed by OpenCode and reviewed in PRs |\n| `src\u002Ftests\u002F` | Deterministic tests for registry, permissions, prompt invariants, and generated Markdown sync |\n\nDo not edit `agents\u002F*.md` by hand. Update the TypeScript source, regenerate Markdown, then review the generated diff.\n\n```bash\npnpm generate:agents\npnpm check\n```\n\nUse this workflow when changing agent behavior:\n\n1. Edit files under `src\u002Fagents\u002F`, `src\u002Fconfig\u002F`, or `src\u002Fadapters\u002F`.\n2. Run `pnpm generate:agents`.\n3. Review the `agents\u002F*.md` diff; this is the final prompt OpenCode receives.\n4. Run `pnpm check`.\n5. Commit both the TypeScript source changes and the generated Markdown changes.\n\nLocal pre-commit hooks run:\n\n```bash\npnpm check:agents\npnpm test\n```\n\n`check:agents` regenerates `agents\u002F*.md` and fails if the generated output differs from the committed files. It does not stage files automatically.\n\n## Agents Reference\n\nThe system uses a **pipeline architecture** where the orchestrator routes work to specialists based on task type and risk level.\n\n### Core Pipeline Agents\n\n| Agent | Purpose | Model | Key Permissions |\n|-------|---------|-------|-----------------|\n| **orchestrator-agent** | Coordinates all specialists, manages lanes and approval gates | `qwen3.6-plus` | read, question; delegates to all other agents |\n| **research-agent** | Investigates code, patterns, dependencies, docs, root causes | `deepseek-v4-flash` | read, websearch, webfetch, question |\n| **planning-agent** | Creates implementation-ready plans with commit structure | `gpt-5.5` (xhigh) | read, edit, delegates to research-agent |\n| **reviewer-agent** | Adversarial stress-testing of plans before implementation | `glm-5.1` | read, git diff\u002Fstatus\u002Frg, question |\n| **implementation-agent** | Executes approved plans step-by-step, production-ready code | `kimi-k2.6` | read, edit, bash (ask), question |\n| **test-fixer-agent** | Diagnoses and repairs failing tests with minimal changes | `deepseek-v4-pro` | read, edit, bash, question |\n| **verifier-agent** | Post-implementation audit against approved plan | `gpt-5.5` (xhigh) | read, git diff\u002Fstatus\u002Frg, question |\n\n### Specialized Agents\n\n| Agent | Purpose | Model | Key Permissions |\n|-------|---------|-------|-----------------|\n| **pr-review-agent** | Reviews GitHub PRs or local branch diffs for correctness, risk, security | `glm-5.1` | read, git commands, gh CLI, question |\n| **prompt-agent** | Refines, critiques, and improves OpenCode agent prompts and workflows | `qwen3.6-plus` | read, question |\n\n### Execution Lanes\n\nThe orchestrator classifies work into three lanes:\n\n| Lane | Use For | Flow |\n|------|---------|------|\n| **Quick Lane** | Small changes, single-file edits, docs, obvious fixes | `research` → `planning` → `implementation` |\n| **Medium Lane** | Multi-file changes, bug fixes, shared abstractions, behavior changes | `research` → `planning` → `review` → `implementation` → `verify` |\n| **Full Lane** | Migrations, auth\u002Fsecurity, data integrity, payments, public APIs | `research` → `planning` → `review` → **approval** → `implementation` → `verify` |\n\n### Debug Lane\n\nFor bug fixing specifically:\n\n1. Triage symptoms\n2. `research-agent` maps suspect files and root-cause candidates\n3. State hypothesis + falsification check\n4. `planning-agent` creates smallest fix plan\n5. `reviewer-agent` critiques if risky\n6. Approval gate\n7. `implementation-agent` applies fix\n8. `test-fixer-agent` if tests break\n9. `verifier-agent` audits result\n\n---\n\n## Commands Guide\n\nCommands trigger the orchestrator with a specific lane configuration.\n\n| Command | Purpose | Lane | When to Use |\n|---------|---------|------|-------------|\n| **quick** | Small, isolated, low-risk changes | Quick lane | Single-file edits, mechanical changes, simple additions |\n| **medium** | Multi-file work needing structure but not full design | Medium lane | Refactors, feature additions touching multiple modules |\n| **full** | Broad, risky, architecture-sensitive work | Full lane | New features, migrations, major refactors, design changes |\n| **debug** | Finding and fixing bugs | Debug lane | Bug reports, regressions, failing tests, unexpected behavior |\n\n### Command Behavior\n\nAll commands run through `orchestrator-agent` as subtasks and follow this pattern:\n\n1. **Intake** — scope the request\n2. **Discovery** — delegate to `research-agent` for codebase mapping\n3. **Planning** — `planning-agent` creates implementation plan saved to `plans\u002F{feature-name}\u002Fplan.md`\n4. **Review** — `reviewer-agent` stress-tests the plan (skipped in quick lane)\n5. **Approval** — `[y\u002FN\u002Fedit]` gate before edits begin\n6. **Execution** — `implementation-agent` follows the plan\n7. **Verification** — `verifier-agent` audits the result (skipped in quick lane for trivial changes)\n\n---\n\n## Configuration Details\n\n### opencode.json\n\n```json\n{\n  \"$schema\": \"https:\u002F\u002Fopencode.ai\u002Fconfig.json\",\n  \"plugin\": [\"@warp-dot-dev\u002Fopencode-warp\"],\n  \"autoupdate\": \"notify\",\n  \"model\": \"opencode-go\u002Fqwen3.6-plus\",\n  \"small_model\": \"opencode-go\u002Fqwen3.5-plus\",\n  \"agent\": {\n    \"orchestrator-agent\": {\n      \"model\": \"opencode-go\u002Fqwen3.6-plus\",\n      \"temperature\": 0.1\n    },\n    \"planning-agent\": {\n      \"model\": \"openai\u002Fgpt-5.5\",\n      \"temperature\": 0.1,\n      \"variant\": \"xhigh\"\n    },\n    \"research-agent\": {\n      \"model\": \"opencode-go\u002Fdeepseek-v4-flash\",\n      \"temperature\": 0.1\n    },\n    \"reviewer-agent\": {\n      \"model\": \"opencode-go\u002Fglm-5.1\",\n      \"temperature\": 0.1\n    },\n    \"pr-review-agent\": {\n      \"model\": \"opencode-go\u002Fglm-5.1\",\n      \"temperature\": 0.1\n    },\n    \"implementation-agent\": {\n      \"model\": \"opencode-go\u002Fkimi-k2.6\",\n      \"temperature\": 0.1\n    },\n    \"test-fixer-agent\": {\n      \"model\": \"opencode-go\u002Fdeepseek-v4-pro\",\n      \"temperature\": 0.1\n    },\n    \"verifier-agent\": {\n      \"model\": \"openai\u002Fgpt-5.5\",\n      \"temperature\": 0.1,\n      \"variant\": \"xhigh\"\n    },\n    \"prompt-agent\": {\n      \"model\": \"opencode-go\u002Fqwen3.6-plus\",\n      \"temperature\": 0.2\n    }\n  },\n  \"permission\": {\n    \"bash\": \"ask\",\n    \"edit\": \"allow\"\n  },\n  \"compaction\": {\n    \"auto\": true,\n    \"prune\": true\n  },\n  \"watcher\": {\n    \"ignore\": [\"node_modules\u002F**\", \"dist\u002F**\", \".git\u002F**\", \"build\u002F**\", \".next\u002F**\", \"coverage\u002F**\"]\n  },\n  \"lsp\": true,\n  \"formatter\": true\n}\n\n```\n\n### Model Assignment Strategy\n\n| Role | Model | Rationale |\n|------|-------|-----------|\n| **Orchestration** | `qwen3.6-plus` | Strong reasoning for routing decisions |\n| **Planning** | `gpt-5.5` (xhigh) | Best-in-class for structured plan generation |\n| **Research** | `deepseek-v4-flash` | Fast, cost-effective codebase exploration |\n| **Review** | `glm-5.1` | Strong adversarial analysis |\n| **Implementation** | `kimi-k2.6` | Efficient code generation |\n| **Test Fixing** | `deepseek-v4-pro` | Deep reasoning for test diagnostics |\n| **Verification** | `gpt-5.5` (xhigh) | Thorough audit capability |\n| **Prompt Engineering** | `qwen3.6-plus` | Good at meta-reasoning about prompts |\n\n### Permission System\n\n- **bash**: `ask` — requires user confirmation before running shell commands\n- **edit**: `allow` — agents can modify files within their defined permissions\n- Individual agents have granular overrides (e.g., `research-agent` cannot edit, `implementation-agent` cannot commit)\n\n### Plugin System\n\nThe `@warp-dot-dev\u002Fopencode-warp` plugin provides enhanced capabilities integrated into the workflow.\n\n### Compaction & Watcher\n\n- **Auto-compaction** with pruning keeps context windows efficient\n- **File watcher** ignores build artifacts, dependencies, and git internals\n\n---\n\n## Usage Examples\n\n### Example 1: Quick Fix\n\n``` md\n\u002Fquick add null check to parseUser in auth.ts\n```\n\nThe orchestrator:\n\n1. Confirms the task is small enough for quick lane\n2. Gathers minimal context via `research-agent`\n3. Creates a tiny plan with `planning-agent`\n4. Skips approval (mechanical change)\n5. Delegates to `implementation-agent`\n6. Reports files changed and validations run\n\n### Example 2: Feature Development\n\n``` md\n\u002Ffull add dark mode toggle to settings page\n```\n\nThe orchestrator:\n\n1. Delegates discovery to `research-agent`\n2. `planning-agent` drafts approach with existing patterns\n3. `reviewer-agent` stress-tests the plan\n4. Presents plan at `plans\u002Fdark-mode-toggle\u002Fplan.md`\n5. Asks `[y\u002FN\u002Fedit]` for approval\n6. Executes tracks via `implementation-agent`\n7. `verifier-agent` audits the result\n\n### Example 3: Debugging\n\n``` md\n\u002Fdebug API returns 500 when creating orders with empty cart\n```\n\nThe orchestrator:\n\n1. Triages: symptom = 500 on empty cart order creation\n2. `research-agent` maps order creation flow, finds suspect validation\n3. Hypothesis: missing null check on cart items array\n4. `planning-agent` creates minimal fix plan\n5. `reviewer-agent` critiques (touches shared order logic)\n6. Approval gate → `[y\u002FN\u002Fedit]`\n7. `implementation-agent` applies fix\n8. `verifier-agent` confirms the fix resolves the 500\n\n### Example 4: PR Review\n\n``` md\n@pr-review-agent review PR #42\n```\n\nThe PR review agent:\n\n1. Collects evidence via `gh pr view`, `gh pr diff`, `gh pr checks`\n2. Reviews across 7 dimensions: correctness, regression, tests, security, performance, maintainability, CI\n3. Returns verdict: `approve` | `comment` | `request changes`\n4. Lists blocking and non-blocking findings with evidence\n\n### Best Practices\n\n1. **Start with the right lane** — use `\u002Fquick` for small changes, escalate to `\u002Fmedium` or `\u002Ffull` as needed\n2. **Let skills auto-activate** — skills trigger based on context; use `@skill-name` for explicit invocation\n3. **Review plans before approving** — the approval gate is your checkpoint\n4. **Trust the pipeline** — research before planning, review before implementation, verify after\n5. **Use caveman mode** for token efficiency when working with large contexts\n","OpenCode Configuration 是一个将 OpenCode 转变为多代理开发系统的配置项目。它通过9个专业代理构建了一个从研究、计划、审查到实施和验证的完整工作流程，并提供了10种技能，涵盖浏览器自动化、前端最佳实践（如Angular、React、NestJS）、测试工具（Jest、Karma、Playwright）等。此外，还支持4种基于车道的命令以不同严格程度路由工作任务，并通过`caveman`技能实现高效的代理间通信压缩。该项目使用TypeScript编写，支持与`@warp-dot-dev\u002Fopencode-warp`插件集成。适用于需要高效协作开发环境及自动化工作者的团队或个人开发者。",2,"2026-06-11 04:07:26","CREATED_QUERY"]