Claude Skills
Curated SkillsRepositoriesNominate
Sign inSign up
© 2026 Claude Skills·Learn more about Agent Skills
Back to repository

t-plan

stale

Thorough planning for complex features using Task-based orchestration. Turn discussions into executable, self-contained implementation plans. USE for multi-file features, architectural changes, unfamiliar tech requiring research. SKIP for quick fixes or single-file changes.

View on GitHub

Marketplace

bengous-plugins

bengous/claude-code-plugins

Plugin

conductor

Repository

bengous/claude-code-plugins
2stars

conductor/skills/t-plan/SKILL.md

Last Verified

February 4, 2026

Install Skill

Select agents to install to:

Scope:
npx add-skill https://github.com/bengous/claude-code-plugins/blob/main/conductor/skills/t-plan/SKILL.md -a claude-code --skill t-plan

Installation paths:

Claude
.claude/skills/t-plan/
Powered by add-skill CLI

Instructions

# T-Plan Skill (Task-Based Orchestration)

## MANDATORY: STOP AND READ THIS FIRST

**You are an ORCHESTRATOR, not an implementer.**

Before doing ANYTHING else, you MUST:

1. **Create the session directory** with `mkdir -p .t-plan/<session-id>`
2. **Write intent.md** capturing the user's request
3. **Create the master task** with TaskCreate

**You MUST NOT:**
- Use Read, Grep, Glob, or WebSearch yourself (that's the EXPLORE subagent's job)
- Research documentation yourself (that's the VALIDATE subagent's job)
- Skip directly to analysis or recommendations

**If you find yourself exploring the codebase before writing intent.md, STOP. You have violated the workflow.**

The user invoked `/t-plan` because they want the FULL orchestrated workflow with artifacts, not ad-hoc help. Follow the steps below EXACTLY.

---

Transform conversations into rock-solid implementation plans using Claude Code's native Task tools for coordination.

## Architecture Overview

```
No hooks required - Task tools handle all coordination natively.

Orchestrator responsibilities:
- Session initialization (mkdir, .gitignore, current.txt pointer)
- Task lifecycle (TaskCreate, TaskUpdate)
- Pre-truncation before subagent dispatch
- Output verification after subagent return
- Retry logic (max 2 attempts)

Subagent responsibilities:
- Read context files (intent.md, explore.md, etc.)
- Write output files directly (explore.md, scout.md, validation-vNNN.json)
- Focus on their assigned task
```

**Reference material** (load when needed):
- Retry logic & parallel execution: See `references/subagent-patterns.md`
- Planning principles & anti-patterns: See `references/planning-principles.md`
- Plan output template: See `references/plan-template.md`

---

## Workflow

```
INTENT -> EXPLORE -> [gate] -> SCOUT -> DRAFT -> VALIDATE -> PLAN
   |          |                   |        |          |         |
 orch.     subagent            subagent  orch.    subagent   orch.+user
```

| Step | Actor | Output |
|----

Validation Details

Front Matter
Required Fields
Valid Name Format
Valid Description
Has Sections
Allowed Tools
Instruction Length:
9562 chars

Issues Found:

  • placeholder_text