メニュー
Use when the user asks "what should I work on", mentions GSD phases, ROADMAP, or .planning/ artifacts. Routes to the correct GSD command for the current project state.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | gsd-workflow |
| description | Use when the user asks "what should I work on", mentions GSD phases, ROADMAP, or .planning/ artifacts. Routes to the correct GSD command for the current project state. |
| description-frequency | on-demand |
| user-invocable | true |
| version | 1.0.0 |
| format | "2025-10-02T00:00:00.000Z" |
| triggers | ["the user asks \"what should I work on\", mentions GSD phases, ROADMAP, or"] |
| updated | "2026-04-25T00:00:00.000Z" |
| status | ACTIVE |
Route user intent to GSD commands. Explicit /gsd: commands bypass routing -- read .claude/commands/gsd/[command].md directly.
When the user provides natural language related to project workflow, route through the GSD Orchestrator's classification pipeline:
When to use the orchestrator: The user says something like "review the requirements," "what's the project status," "let's verify phase 3," or any natural language that maps to a GSD lifecycle action but doesn't use explicit slash commands.
When to bypass it: The user types an explicit GSD slash command. Execute directly.
When confidence is low: Ask the user to clarify rather than guessing. A wrong routing is worse than a clarifying question.
The top-10 most common routes:
| User Says | Route To |
|---|---|
| "what should I work on" | /gsd:progress |
| "continue where I left off" | /gsd:progress |
| "build phase X" | /gsd:execute-phase X |
| "plan the next phase" | /gsd:plan-phase N |
| "discuss how phase X should work" | /gsd:discuss-phase N |
| "something's broken" | /gsd:debug |
| "quick fix / small task" | /gsd:quick |
| "verify phase X" | /gsd:verify-work X |
| "add a phase" | /gsd:add-phase |
| "start new project" | /gsd:new-project |
For full routing tables including skill-creator actions, see references/command-routing.md.
Always suggest GSD when:
Suggest with explanation:
This looks like a good candidate for `/gsd:plan-phase N` -- it'll break this
down into atomic tasks with verification criteria. Want me to run that,
or would you prefer to dive in directly?
Suggest skill-creator when:
Suggest naturally:
I've noticed we run the same lint -> test -> fix cycle after every code change.
Want me to capture this as a skill so it happens automatically?
Respect user override when:
Acknowledge gracefully:
Got it -- working on this directly. If it grows in scope, we can always
capture it in a plan retroactively with `/gsd:quick`.
Gently push back when:
Push back helpfully:
Before I make changes, let me check `/gsd:progress` -- there might be
existing plans that touch this area. One moment...
| Marker | Meaning | Claude Action |
|---|---|---|
## > Next Up | Next command to run | Read the instruction and execute it |
/clear first -> | Context window full, needs reset | Tell user to run /clear, then continue |
Ready to build | Planning complete, execution ready | Proceed to /gsd:plan-phase or /gsd:execute-phase |
ROADMAP CREATED | Roadmap agent finished | Review output, proceed to next phase |
PLAN CREATED | Planning agent finished | Proceed to execution |
Standard Cycle:
/gsd:plan-phase N -> /clear -> /gsd:execute-phase N -> /gsd:verify-work N
Fresh Session Recovery:
/gsd:progress (or) /gsd:resume-work
Mid-Work Context Reset:
/gsd:pause-work -> /clear -> /gsd:resume-work
Debugging Flow:
/gsd:debug "description" -> investigate -> /clear -> /gsd:debug (resume)
For phase transition details and response patterns, see references/phase-behavior.md.
Detection: Read .planning/config.json. If "mode": "yolo", autonomous operation is active.
6-Step Autonomous Protocol:
/clear requirements -- inform user when context reset is neededFor example recognition and edge cases, see references/yolo-mode.md.
To run a GSD command, read the command file from .claude/commands/gsd/[command].md and follow its process. The command file contains:
<objective> -- What the command achieves<process> -- Step-by-step instructions to follow<success_criteria> -- How to know it's completeCritical: Don't just acknowledge GSD output -- act on it.
Before executing any GSD phase, load relevant generated skills:
.claude/commands/ for project-level skills~/.claude/commands/ for user-level skillsCritical: When forking subagent contexts for GSD phases (execute-phase, verify-work), include relevant skills in the subagent's context. Clean context means free of stale conversation history -- not free of learned knowledge.
/gsd:plan-phaseExtract creative intent from images into executable build specs. Activates on images + build intent, "image to mission", "i2m", or capturing visual energy in code/design.
Unit Circle Laboratory + sc-dev-team autonomous mission control. Provides human-in-the-loop automation layer, pipeline management, context lifecycle, stuck-state prevention. UC re-execution series and dev branch milestones. Adapted from uc-lab pattern for code plan/execute/verify/complete. Auto-activates during UC milestone work; trigger: user says 'bring up the dev team'.
Generate intelligence briefings for the planning dashboard. Use this skill whenever a request file appears in `.planning/console/inbox/pending/` whose `type` field starts with `intelligence.` (refresh_briefing, triage_finding, snapshot_diff, investigate_section, dismiss_finding). The skill reads the per-project KB at `.gsd/intelligence/intelligence.db`, synthesizes a briefing with a causal hypothesis + acknowledged uncertainty + confidence label and ranked moves, then writes the result back to the KB. Always trigger this skill for these request types — do not generate briefings manually.
Sketched Isotropic Gaussian Regularization primitive. Scalar loss matching the embedding distribution to a standard-normal target via Cramér-Wold slicing and the Epps-Pulley empirical characteristic function test. Port of rbalestr-lab/lejepa (MIT). Default-off in v1.49.571.
Adversarial spec-compliance PR review — cross-references diffs against approved specs, verifies runtime claims against source, detects competing PRs, audits scope/convention compliance. Use before merging.
Pick among candidate outputs (code, configs, plans) by running them on diverse inputs and clustering by behavioural fingerprint, rather than by textual aggregation or log-probability. Activates when an executor returns multiple plausible candidates that need disambiguation, when output-majority voting would be the default choice, or when reviewing generated code that has not yet been validated. The 2026 evidence (Semantic Voting, arxiv 2605.08680v1) is that any execution-based selector dominates output-majority voting by 19-52pp; sketch-generated inputs beat random fuzz by 11.3pp. Triggers: "pick the best candidate", "majority vote on code", "select from N samples", "validate the generated output", "behavioural verification".