Multi-agent ensemble planning with thunk CLI. Use when discussing implementation plans, task planning, or when user mentions thunk sessions. Knows thunk CLI syntax and workflow.
Multi-agent ensemble planning with thunk CLI. Use when discussing implementation plans, task planning, or when user mentions thunk sessions. Knows thunk CLI syntax and workflow.
/plugin marketplace add gbasin/thunk/plugin install thunk@thunkThis skill inherits all available tools. When active, it can use any tool Claude has access to.
If the thunk command is not available, install it first:
cd /path/to/thunk/repo # where pyproject.toml lives
uv venv && source .venv/bin/activate
uv pip install -e .
Or in an existing venv: source .venv/bin/activate && uv pip install -e .
Thunk orchestrates multiple AI agents (Claude Code + OpenAI Codex) to create implementation plans through iterative refinement.
| Command | Purpose |
|---|---|
thunk init "task" | Start new planning session |
thunk wait --session <id> | Block until turn complete |
thunk status --session <id> | Check progress |
thunk continue --session <id> | Start next turn after edits |
thunk approve --session <id> | Lock plan as final |
thunk list | List all sessions |
thunk clean --session <id> | Remove session |
thunk diff --session <id> | Show changes between turns |
init → wait → [user edits] → continue → wait → ... → approve
.thunk/sessions/<id>/turns/NNN.mdinitializing - Session just createddrafting - Agents creating initial draftspeer_review - Agents reviewing each other's worksynthesizing - Combining into unified planuser_review - Waiting for human feedbackapproved - Plan locked and final## Questions
### Q1: [Question needing user input]
**Context:** Why this matters
**Answer:** [User fills this in]
## Summary
[2-3 sentence overview]
## Tasks
- [ ] **Task 1**: Description
- **Files:** `path/file.py` (create|modify)
- **Rationale:** Why needed
- **Dependencies:** none | Task N
## Risks
- **Risk name** (severity: high|medium|low)
- **Mitigation:** How to address
## Alternatives Considered
- **Alternative**: Rejected because [reason]
.thunk/sessions/swift-river/ # Human-friendly session ID
├── meta.yaml # Task, created_at
├── state.yaml # turn, phase, agent_plan_ids
├── bold-peak.md # Agent's working plan (opaque name)
├── calm-forest.md # Another agent's plan
├── turns/
│ ├── 001.md # Turn 1 synthesis (user edits this)
│ └── 002.md # Turn 2
├── agents/
│ ├── opus/cli_session_id.txt
│ ├── codex/cli_session_id.txt
│ └── turn-001/
│ ├── opus-draft.md
│ └── opus-final.md
└── PLAN.md # Symlink to approved turn
Agents preserve context across turns:
--resume <session_id>codex resume <thread_id>First turn exploration (reading files, understanding patterns) is remembered in subsequent turns.
This skill should be used when the user asks to "create a hookify rule", "write a hook rule", "configure hookify", "add a hookify rule", or needs guidance on hookify rule syntax and patterns.