メニュー
Deploy, configure, and maintain the Solo-Code multi-agent harness across Copilot, Claude, Gemini, and Kilo. Use when: deploy harness, setup copilot, setup claude, setup gemini, setup kilo, configure AI agents, install AI harness, solo-code setup, initialize project.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | solo-code-harness |
| description | Deploy, configure, and maintain the Solo-Code multi-agent harness across Copilot, Claude, Gemini, and Kilo. Use when: deploy harness, setup copilot, setup claude, setup gemini, setup kilo, configure AI agents, install AI harness, solo-code setup, initialize project. |
| license | MIT |
You are an expert on the Solo-Code Harness framework. This project provides discipline, safety, and consistency rules for AI coding agents across four platforms: GitHub Copilot (VS Code), Claude Code, Gemini, and Kilo.
The harness lives in four platform-specific directories at the project root:
| Platform | Directory | Auto-loads |
|---|---|---|
| GitHub Copilot / Cursor | .github/ | copilot-instructions.md + PreToolUse hooks |
| Claude Code | .claude/ | CLAUDE.md |
| Gemini | .gemini/antigravity/ | AGENTS.md + skills |
| Kilo | .kilo/ + kilo.json + AGENTS.md | AGENTS.md, command/, agent/, skill/ |
Each harness includes: rulebook, skills (8-9), agents (2), memory, automation scripts.
Copy this entire repository's root contents (excluding .git/, .github/workflows/, etc.) into a target project. Each AI tool auto-discovers its respective config directory.
Copy to target project:
AGENTS.md
kilo.json
.kilo/
Copy to target project:
.github/
.vscode/
The .kilo/ directory is auto-discovered by Kilo. No additional config needed.
.kilo/agent/)| Agent | Mode | Purpose |
|---|---|---|
orchestrator.md | subagent | Breaks down complex tasks, coordinates parallel sub-agents |
solo-code-engineer.md | subagent | Template agent for project customization |
.kilo/command/)| Command | Purpose |
|---|---|
/plan | Implementation planning with affected files, tasks, risks |
/verify | Prove code works by building, running, and testing |
/brainstorm | Structured idea exploration with 3+ options |
/debug | Evidence-based root cause analysis before fixes |
/test | Test generation and execution (AAA pattern) |
/create | Feature scaffolding from scratch |
/enhance | Safe refactoring of existing code |
/remember | Save project conventions to persistent memory |
.kilo/skill/)| Skill | When to load |
|---|---|
code-review-expert | Reviewing code, PRs, diffs; security audits |
file-editor-pro | Editing files, refactoring, precise changes |
git-workflow-master | Committing, pushing, PRs, git safety |
permission-guard | Destructive operations, credentials, config |
systematic-debugging | Debugging, errors, unexpected behavior |
brainstorming | Designing features, architecture decisions |
testing-patterns | Writing tests, TDD, coverage |
api-patterns | API design (REST/GraphQL/tRPC) |
.kilo/agent/solo-code-engineer.md with your project's tech stack, architecture map, and conventions.kilo/memory/project-conventions.md with your project's specific ruleskilo.json permissions if your project needs different bash command accessCreate .kilo/skill/<name>/SKILL.md with YAML frontmatter:
name: your-skill-name
description: "When to use this skill"
The skill body can use any markdown format Kilo supports.
Create .kilo/command/<name>.md with YAML frontmatter:
description: What the command does
subtask: true
The command name is the filename (minus .md), invoked via /name.
The harness uses three layers of protection:
kilo.json) — Deny patterns evaluated before allow patterns. Destructive git/rm/DB/Windows-PowerShell operations blocked at the bash level.AGENTS.md) — Rules instructing the AI: no destructive commands without permission, scan for secrets before commit, no runtime bypass of bash permissions.permission-guard halts execution for dangerous operations; git-workflow-master enforces secret scanning with security_scan.py before commits.When editing shared concepts (behavior rules, security rules, project conventions), update ALL four harnesses:
.github/copilot-instructions.md.claude/CLAUDE.md.gemini/antigravity/AGENTS.mdAGENTS.md (Kilo root)Bắt buộc khai báo độ phức tạp Big-O và cấu trúc dữ liệu trước khi viết thuật toán.
Guides stable API and interface design. Use when designing APIs, module boundaries, or any public interface. Use when creating REST or GraphQL endpoints, defining type contracts between modules, or establishing boundaries between frontend and backend.
Prevent commits that skip pre-commit hooks (--no-verify, -n). Auto-loads when git commit with bypass flags is attempted. Triggers: commit --no-verify, git commit -n, git push --no-verify.
Explore ideas before implementation, design features, make architecture decisions, and validate approaches. Use when: brainstorm, explore options, design, what approach, how should we, ideas, before building.
Tests in real browsers via Chrome DevTools MCP. Use when building or debugging anything that runs in a browser. Use when you need to inspect the DOM, capture console errors, analyze network requests, profile performance, or verify visual output with real runtime data. Requires the chrome-devtools MCP server to be configured.
Automates CI/CD pipeline setup. Use when setting up or modifying build and deployment pipelines.