菜单
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.
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
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.
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.
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 |
| version | 0.3.0 |
| description | 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. |
Check for .planning/ directory in the project root. This single check determines routing.
| Scenario | .planning/ exists? | System | Entry Point |
|---|---|---|---|
| New project (greenfield) | No → creates it | GSD | /gsd:new-project |
| Brownfield onboarding | No → creates it | GSD | /gsd:map-codebase then /gsd:new-project |
| New feature in GSD project | Yes | GSD | /gsd:discuss-phase → /gsd:spec-phase → /gsd:plan-phase → /gsd:execute-phase |
| AI/LLM integration phase | Yes | GSD | /gsd:ai-integration-phase → /gsd:plan-phase → /gsd:execute-phase → /gsd:eval-review |
| Quick task in GSD project | Yes | GSD | /gsd:quick |
| Trivial fix in GSD project | Yes | GSD | /gsd:fast (inline, no subagents) |
| Standalone feature (no GSD) | No | SP | REQUIRED: brainstorming → writing-plans → subagent-driven-development |
| PRD-driven feature | Either | SP → GSD | write-a-prd → prd-to-issues → /gsd:new-project or /gsd:add-phase |
| Bug fix (any context) | Either | SP first | REQUIRED: systematic-debugging ALWAYS first, then route fix through appropriate system |
| Code review (send) | Either | SP | REQUIRED: requesting-code-review (fills GSD's missing review step) |
| Code review (receive) | Either | SP | receiving-code-review — verify feedback before implementing blindly |
| Stress-test a plan | Either | SP | grill-me — interview until all ambiguity is resolved |
| Architecture improvement | Either | SP | improve-codebase-architecture — surface friction, propose refactors as RFC issues |
| Ship / PR creation | Yes | GSD | /gsd:ship or /gsd:pr-branch (filters .planning/ commits) |
| Ship (standalone) | No | SP | finishing-a-development-branch |
| Resume work | Yes | GSD | /gsd:resume-work |
| Resume work | No | SP | Re-read plan files in docs/plans/ |
| Scenario | Entry Point | Notes |
|---|---|---|
| Spec refinement | /gsd:spec-phase | Socratic clarification of WHAT a phase delivers — produces SPEC.md before planning |
| AI/LLM design contract | /gsd:ai-integration-phase | Framework selection, implementation guidance, evaluation strategy → AI-SPEC.md |
| AI eval audit | /gsd:eval-review | Retroactive audit of AI phase evaluation coverage |
| Spike (feasibility test) | /gsd:spike → /gsd:spike-wrap-up | Throwaway experiment to validate approach before committing |
| Sketch (UI mockup) | /gsd:sketch → /gsd:sketch-wrap-up | Rapid HTML mockups with multi-variant exploration |
| UI/frontend phase | /gsd:ui-phase → /gsd:ui-review | Design contract (UI-SPEC.md) before implementation |
| Security audit | /gsd:secure-phase | Verify threat mitigations post-implementation |
| Codebase intelligence | /gsd:intel | Query, refresh, or inspect .planning/intel/ |
| Code review + auto-fix | /gsd:code-review → /gsd:code-review-fix | Two-step: find issues, then fix atomically |
| Autonomous execution | /gsd:autonomous | Runs all remaining phases: discuss → plan → execute |
| Parallel workstreams | /gsd:workstreams | Manage concurrent work across features |
| Persistent debugging | /gsd:debug | 4-phase debugging with knowledge base persistence |
| Phase assumptions | /gsd:list-phase-assumptions | Surface Claude's assumptions before planning |
| Nyquist validation | /gsd:validate-phase | Audit and fill validation gaps for a completed phase |
| Milestone audit | /gsd:audit-milestone → /gsd:plan-milestone-gaps | Verify intent delivered, fill gaps |
| Ship / PR creation | /gsd:ship | Create PR, run review, prepare for merge |
| Clean PR branch | /gsd:pr-branch | Filter out .planning/ commits for clean code review |
| Safe revert | /gsd:undo | Roll back phase or plan commits using phase manifest |
| Project statistics | /gsd:stats | Phases, plans, git metrics, timeline |
| Session report | /gsd:session-report | Token usage estimates and work summary |
| Docs generation | /gsd:docs-update | Generate/update project docs verified against codebase |
| Changelog | changelog-generator (SP) | Auto-generate user-facing changelogs from git commits |
| Diagnostics | /gsd:health | Diagnose and repair .planning/ directory issues |
| Post-mortem | /gsd:forensics | Investigate failed GSD workflows |
| Learnings extraction | /gsd:extract_learnings | Extract decisions, patterns, surprises from completed phases |
| Scenario | Entry Point | Notes |
|---|---|---|
| Quick note | /gsd:note | Zero-friction idea capture — append, list, or promote |
| Add todo | /gsd:add-todo | Capture task from conversation context |
| Check todos | /gsd:check-todos | List pending todos and select one to work |
| Backlog idea | /gsd:add-backlog | Park an idea in 999.x numbering |
| Review backlog | /gsd:review-backlog | Promote backlog items to active milestone |
| Plant seed | /gsd:plant-seed | Forward-looking idea with trigger conditions |
| Explore idea | /gsd:explore | Socratic ideation — think before committing |
| Scenario | Entry Point | Notes |
|---|---|---|
| Test web app | webapp-testing (SP) | Playwright-based: screenshots, form fills, link checks |
| Full browser automation | playwright-skill (SP) | Auto-detect dev servers, write test scripts |
GSD orchestrates WHAT to build and in what order. Superpowers enforce HOW to build each piece correctly.
| Command | Purpose |
|---|---|
/torque:status | Where am I? Single view across GSD + SP state |
/torque:check | Verify GSD + SP installation and detect misconfigurations |
/torque:next | Smart next-action — reads state, suggests one command |
/torque:migrate | Upgrade path to Meridian |
.planning/phases/, SP uses docs/plans/. Never mix.discuss-phase for GSD work, brainstorming for standalone only. Never both.execute-phase OR SP subagent-driven-development. Never nested.systematic-debugging runs first for all bugs. GSD deviation Rule 1 covers only incidental executor bugs.requesting-code-review after GSD verify-work, before merge.receiving-code-review to verify suggestions before implementing blindly./gsd:ui-phase to produce UI-SPEC.md before implementation, then /gsd:ui-review after./gsd:ai-integration-phase to produce AI-SPEC.md before implementation, then /gsd:eval-review after./gsd:spec-phase between discuss and plan to lock falsifiable requirements./gsd:spike before discuss-phase. Wrap findings with /gsd:spike-wrap-up./gsd:ship or /gsd:pr-branch to create PRs that exclude .planning/ commits from code review./gsd:intel to inform future phases.brainstorming inside a GSD phase (use discuss-phase)docs/plans/ in a GSD project (use .planning/phases/)subagent-driven-development inside GSD execute-phasesystematic-debugging for bugs in GSD projects.planning/ manually (use /gsd:new-project)/gsd:quick for multi-phase work (use full pipeline)discuss-phase AND brainstorming (pick one based on .planning/ existence)receiving-code-review)/gsd:spec-phase between discuss and plan)/gsd:ai-integration-phase first).planning/ commits in PRs (use /gsd:pr-branch to filter them out)/gsd:spike for uncertain feasibility)Trivial (typo, config) → /gsd:fast
Small (< 30 min, < 5 files) → /gsd:quick
Medium (single phase) → discuss → spec → plan → execute
Large (multi-phase) → new-project or add-phase pipeline
Uncertain feasibility → /gsd:spike first, then route by size
AI/LLM integration → /gsd:ai-integration-phase → plan → execute → eval-review
Autonomous (hands-off) → /gsd:autonomous
For detailed workflows and conflict resolution, load files from references/. Workflow guides are in references/workflows.md.