菜单
Use when a feature touches 4 or more files, involves multiple components, or has subtasks that could run in parallel.
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
Task management CLI for tracking and managing feature subtasks with status, dependencies, and validation
Use when coding standards, security patterns, or project conventions need to be discovered before implementation begins.
Install context files from registry. Use when user runs /install-context, says "install context", "setup context", or when context is missing and the user needs to get started.
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes
Use when the task involves an external library or package and current API docs are needed before writing code.
Use before any implementation — understands the request, discovers project context, and proposes a concise plan for user approval before writing any code.
| name | task-breakdown |
| description | Use when a feature touches 4 or more files, involves multiple components, or has subtasks that could run in parallel. |
| context | fork |
| agent | task-manager |
Break down complex features into atomic, verifiable subtasks with dependency tracking. Each subtask gets its own JSON file with clear acceptance criteria and deliverables.
Announce at start: "I'm using the task-breakdown skill to create an execution plan for [feature]."
Identify these elements:
Write .tmp/tasks/{feature}/task.json:
{
"id": "jwt-auth",
"name": "JWT Authentication System",
"status": "active",
"objective": "Implement JWT-based authentication with refresh tokens",
"context_files": [
".opencode/context/core/standards/code-quality.md",
".opencode/context/core/standards/security-patterns.md"
],
"reference_files": [
"src/middleware/auth.middleware.ts"
],
"exit_criteria": [
"All tests passing",
"JWT tokens signed with RS256"
],
"subtask_count": 3,
"completed_count": 0,
"created_at": "2026-02-16T02:00:00Z"
}
Rules:
context_files: standards/conventions ONLYreference_files: project source files ONLYWrite .tmp/tasks/{feature}/subtask_01.json, subtask_02.json, etc:
{
"id": "jwt-auth-01",
"seq": "01",
"title": "Create JWT service with token generation",
"status": "pending",
"depends_on": [],
"parallel": true,
"suggested_agent": "CoderAgent",
"context_files": [
".opencode/context/core/standards/security-patterns.md"
],
"reference_files": [],
"acceptance_criteria": [
"JWT tokens signed with RS256 algorithm",
"Access tokens expire in 15 minutes"
],
"deliverables": [
"src/auth/jwt.service.ts",
"src/auth/jwt.service.test.ts"
]
}
Rules:
depends_on arrayparallel: true for isolated workVerify:
## Tasks Created
Location: .tmp/tasks/jwt-auth/
Files: task.json + 3 subtasks
Subtasks:
- 01: Create JWT service (parallel: true, agent: CoderAgent)
- 02: Create password hashing util (parallel: true, agent: CoderAgent)
- 03: Integrate middleware (parallel: false, agent: CoderAgent)
Next Steps:
- Execute subtasks in dependency order
- Tasks 01 and 02 can run in parallel
- Task 03 depends on completion of 01 and 02
If you think any of these, STOP and re-read this skill:
| Excuse | Reality |
|---|---|
| "It's only 3-4 files, I don't need a breakdown" | 3-4 files = multiple subtasks with dependencies. Skipping tracking means losing progress on failure. |
| "I'll track it in my head" | Subagents don't share memory. JSON files are the only reliable state. |
| "The tasks are obvious, no need to document them" | Obvious tasks still need acceptance criteria. "Done" without binary criteria is not done. |
| "Parallel execution isn't worth it for this" | Parallel tasks cut execution time in half. The JSON overhead is 2 minutes. The time saving is 20+. |
context_files = standards ONLY, reference_files = source code ONLYparallel: trueTask: Break down this feature into atomic subtasks: $ARGUMENTS