菜单
Create and manage subagent entities in the agent-harness workspace. Invoke this skill when a user wants to define, scaffold, apply, or remove a subagent across Claude Code, OpenAI Codex, or GitHub Copilot.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Create and manage MCP server config entities in the agent-harness workspace.
Create and manage MCP server config entities in the agent-harness workspace.
Create and manage lifecycle hook entities in the agent-harness workspace. Covers canonical event authoring, provider projection (Claude Code, GitHub Copilot, OpenAI Codex CLI, Cursor), mode behavior, handler types, the full event/provider support matrix, copy-paste recipes, and common diagnostics.
Create and manage lifecycle hook entities in the agent-harness workspace. Covers canonical event authoring, provider projection (Claude Code, GitHub Copilot, OpenAI Codex CLI, Cursor), mode behavior, handler types, the full event/provider support matrix, copy-paste recipes, and common diagnostics.
Run a multi-round critical review loop using specialized subagents. Use when the user asks to review changes, harden code, or ensure quality before merge. Spawns parallel review agents across security, core-logic, tests, CLI integration, and API surface, then iterates fix-and-re-review rounds until all agents report clean.
Create and manage skill entities in the agent-harness workspace. Use when the user wants to add a new skill, understand how skills work across providers, or generate provider skill artifacts from a canonical source.
| name | harness-subagent |
| description | Create and manage subagent entities in the agent-harness workspace. Invoke this skill when a user wants to define, scaffold, apply, or remove a subagent across Claude Code, OpenAI Codex, or GitHub Copilot. |
Use this skill to create and manage subagent entities in the agent-harness workspace. A subagent is a specialized agent definition — with a name, description, optional tool restrictions, and an optional model override — that harness renders into provider-native agent configuration files for every enabled provider.
A subagent entity is a single canonical Markdown source file under .harness/src/subagents/<subagent-id>.md. It contains:
name and descriptionHarness renders one output artifact per enabled provider from that single source file. You edit the source once; all provider files stay in sync.
| Item | Value |
|---|---|
| Output path | .claude/agents/<subagent-id>.md |
| Format | Markdown with YAML frontmatter |
| Required frontmatter | name, description |
| Optional frontmatter | tools (string or string[]), model |
Claude Code loads agent files from .claude/agents/ at session start. Each file's description field tells Claude when to delegate to that agent. Claude routes tasks automatically, or the user can invoke an agent explicitly with @agent-<name> or the /agents command. Supported model values: sonnet, opus, haiku, a full model ID such as claude-sonnet-4-6, or inherit (default). Subagents run in isolated context windows and cannot spawn further subagents.
Official docs: https://code.claude.com/docs/en/sub-agents
| Item | Value |
|---|---|
| Output path | .codex/config.toml |
| Format | TOML, section [agents.<subagent-id>] |
| Required fields | description, developer_instructions (body) |
| Optional fields | model, model_reasoning_effort, nickname_candidates |
Harness writes all enabled subagents into .codex/config.toml under [agents.<id>] entries. The description drives when Codex routes to that agent role; the body becomes developer_instructions. Codex also supports standalone TOML agent files in .codex/agents/ (outside of harness management), which are auto-discovered and take precedence over same-named inline entries.
Official docs: https://developers.openai.com/codex/config-reference
| Item | Value |
|---|---|
| Output path | .github/agents/<subagent-id>.agent.md |
| Format | Markdown with YAML frontmatter |
| Required frontmatter | description |
| Optional frontmatter | name, tools (string[]), model, handoffs, mcp-servers, target, disable-model-invocation, user-invocable |
Copilot discovers agent files in .github/agents/ at the repository level. The handoffs field is Copilot-specific and is supported in VS Code agent mode: it defines routing buttons that let users switch to another agent mid-conversation, optionally with a pre-filled prompt. Note: handoffs are currently not supported by the Copilot coding agent on GitHub.com — they are silently ignored there. Example handoff entry (VS Code only):
handoffs:
- label: "Start Implementation"
agent: implementation
prompt: "Now implement the plan outlined above."
send: true
When send: true is set, the handoff prompt submits automatically.
Official docs: https://docs.github.com/en/copilot/reference/custom-agents-configuration
| Item | Value |
|---|---|
| Output path | .cursor/agents/<subagent-id>.md |
| Format | Markdown with YAML frontmatter |
| Required frontmatter | name, description |
| Optional frontmatter | model, readonly, is_background |
Cursor loads agent files from .cursor/agents/ (project) or ~/.cursor/agents/ (user). For compatibility, it also scans .claude/agents/ and .codex/agents/ (project-level) and their user-level equivalents. Project subagents take precedence when names conflict; .cursor/ takes precedence over .claude/ or .codex/.
The description field determines when the agent delegates to a subagent. Subagents can be invoked explicitly via /name syntax or automatically based on task complexity. The model field supports fast, inherit (default), or a specific model ID. When readonly: true, the subagent runs with restricted write permissions.
Cursor includes three built-in subagents (explore, bash, browser) that run automatically. Custom subagents cannot spawn further subagents (single-level only).
Official docs: https://docs.cursor.com/agent/subagents
| Field | Claude Code | Codex CLI | GitHub Copilot | Cursor |
|---|---|---|---|---|
name | required | via TOML key | optional | required |
description | required | required | required | required |
tools | string or string[] | — | string[] | — |
model | optional | optional | optional | optional (fast, inherit, or model ID) |
readonly | — | — | — | optional (boolean) |
is_background | — | — | — | optional (boolean) |
developer_instructions | — | required (body) | — | — |
handoffs | — | — | optional (VS Code only) | — |
Provider-specific options (model, tools, handoffs) are set through override sidecar files, not in the canonical source frontmatter — see the Override sidecars section below.
---
name: <human-readable display name>
description: <when this agent should be used — used by all providers>
---
You are a [role description]. When invoked:
1. [Step one]
2. [Step two]
3. [Step three]
Focus on [specific domain]. Do not [out-of-scope actions].
Only name and description are required in the canonical frontmatter. The body becomes the agent's system prompt for all providers.
To set provider-specific options (model, tools, handoffs), create or edit the generated override YAML files:
.harness/src/subagents/<subagent-id>.overrides.claude.yaml.harness/src/subagents/<subagent-id>.overrides.codex.yaml.harness/src/subagents/<subagent-id>.overrides.copilot.yamlExample Claude override sidecar:
version: 1
options:
model: haiku
tools:
- Read
- Grep
- Glob
Example Copilot override sidecar:
version: 1
options:
model: gpt-4o
tools:
- search
- fetch
handoffs:
- label: "Hand off to planner"
agent: planner
prompt: "Now create a plan based on the research above."
# Scaffold a new subagent source file and register it in manifest.json
npx harness add subagent <subagent-id>
# Preview what files will be generated (dry run)
npx harness plan
# Write all provider artifacts from current canonical sources
npx harness apply
# Watch sources and auto-apply on changes
npx harness watch
# Remove a subagent entity and its source files
npx harness remove subagent <subagent-id>
# Remove entity but keep source files on disk
npx harness remove subagent <subagent-id> --no-delete-source
Source file: .harness/src/subagents/code-reviewer.md
---
name: code-reviewer
description: Expert code review specialist. Reviews code for quality, security, and maintainability. Use proactively after writing or modifying code.
---
You are a senior code reviewer ensuring high standards of quality and security.
When invoked:
1. Run git diff to see recent changes
2. Focus on modified files
3. Begin review immediately without asking for clarification
Review checklist:
- Code is clear and readable
- Functions and variables are well-named
- No duplicated logic
- Proper error handling and input validation
- No exposed secrets or API keys
- Adequate test coverage
- Performance considerations addressed
Provide feedback organized by priority:
- Critical issues (must fix before merging)
- Warnings (should fix)
- Suggestions (consider improving)
Include specific examples of how to fix each issue. Do not modify files yourself.
Override sidecar for Claude (.harness/src/subagents/code-reviewer.overrides.claude.yaml):
version: 1
options:
model: sonnet
tools:
- Read
- Grep
- Glob
- Bash
After running npx harness apply, this produces:
.claude/agents/code-reviewer.md — used by Claude Code.codex/config.toml under [agents.code-reviewer] — used by Codex.github/agents/code-reviewer.agent.md — used by GitHub Copilot# 1. Scaffold the subagent source
npx harness add subagent code-reviewer
# 2. Edit the source file
# .harness/src/subagents/code-reviewer.md
# 3. (Optional) Edit provider-specific overrides
# .harness/src/subagents/code-reviewer.overrides.claude.yaml
# 4. Preview the plan
npx harness plan
# 5. Generate provider artifacts
npx harness apply
Check .claude/agents/, .codex/config.toml, and .github/agents/ to verify the rendered output.