一键导入
torque-next
Determine the single best next action based on current project state. Reads GSD state, SP plans, and git context to suggest exactly one command to run.
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。
菜单
Determine the single best next action based on current project state. Reads GSD state, SP plans, and git context to suggest exactly one command to run.
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
Torque: a Claude Code skill that converts GSD (project management) and Superpowers (engineering discipline) into one coherent workflow. Prevents double-planning, conflicting subagent models, and overlapping design capture by providing a single routing decision at task start.
Verify GSD and Superpowers installation, detect version mismatches, and flag common misconfigurations. Run this when things aren't working as expected.
Evaluate whether to upgrade from Torque (GSD + Superpowers) to Meridian, and guide the migration if ready. Use when projects outgrow the routing model.
Show current project state across GSD and Superpowers systems. Reads .planning/, docs/plans/, and git state to give a unified view of where you are and what to do next.
| name | torque:next |
| description | Determine the single best next action based on current project state. Reads GSD state, SP plans, and git context to suggest exactly one command to run. |
Analyze project state and recommend exactly ONE next command. Think of this as a GPS for your development workflow.
Same as /torque:status — read .planning/, docs/plans/, git state. But don't output the full status. Focus on determining the current position.
Work through this tree top-to-bottom. First match wins.
1. Is there an active bug or failing test?
→ `systematic-debugging`
Why: Bugs block everything. Fix first.
2. Does .planning/ exist?
├─ YES (GSD project):
│ a. Is there a HANDOFF.md in the active phase?
│ → `/gsd:resume-work`
│ Why: Paused work should be resumed before starting new work.
│
│ b. Is STATE.md showing a phase in "executing" with incomplete plans?
│ → `/gsd:execute-phase` (continue)
│ Why: Finish what you started.
│
│ c. Is a phase complete but not verified?
│ → `/gsd:verify-work`
│ Why: Verification before review.
│
│ d. Is a phase verified but not code-reviewed?
│ → `requesting-code-review`
│ Why: GSD has no review step — SP fills the gap.
│
│ e. Is a phase reviewed but not shipped?
│ → `/gsd:ship` or `/gsd:pr-branch`
│ Why: Ship cleanly — filters .planning/ commits from PR.
│
│ f. Are all phases in the current milestone complete?
│ → `/gsd:audit-milestone` then `/gsd:complete-milestone`
│ Why: Close out the milestone properly.
│
│ g. Is the next phase planned but not started?
│ → `/gsd:execute-phase`
│ Why: Plans exist, execute them.
│
│ h. Is the next phase discussed and spec'd but not planned?
│ → `/gsd:plan-phase`
│ Why: Context and spec exist, create plans.
│
│ h2. Is the next phase discussed but not spec'd (ambiguous scope)?
│ → `/gsd:spec-phase`
│ Why: Lock falsifiable requirements before planning.
│
│ i. Is there a next phase that hasn't been discussed?
│ → `/gsd:discuss-phase`
│ Why: Gather context before planning.
│
│ i2. Is this an AI/LLM phase that hasn't been designed?
│ → `/gsd:ai-integration-phase`
│ Why: Produce AI-SPEC.md with framework + eval strategy.
│
│ j. No more phases in roadmap?
│ → `/gsd:new-milestone` or project complete
│
└─ NO (no .planning/):
a. Does docs/plans/ exist with incomplete plans?
→ Re-read plans, continue execution
Why: Pick up where you left off.
b. No workflow state at all?
→ Ask: "New project or standalone feature?"
→ New project: `/gsd:new-project`
→ Existing codebase: `/gsd:map-codebase`
→ Quick standalone: `brainstorming`
## Next Action
→ `/gsd:plan-phase`
**Why**: Phase 3 ("Payment integration") has CONTEXT.md from discuss-phase
but no PLAN.md files yet. Planning is the next step before execution.
**Current position**: Milestone 1, Phase 3 of 7
**What this does**: Creates executable PLAN.md files with task breakdown,
dependencies, and wave assignments for parallel execution.
Keep it to ONE recommendation. If truly ambiguous, show at most two options with a clear default.