メニュー
Budget-weighted prompt execution. depth_probe frames thinking. plan_prompt budgets effort. Per-phase pauses enable midstream metathinking. audit_probe catches gaps.
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 | exec-opt |
| description | Budget-weighted prompt execution. depth_probe frames thinking. plan_prompt budgets effort. Per-phase pauses enable midstream metathinking. audit_probe catches gaps. |
| argument-hint | [question-or-task] [--lite] |
| ultrathink | true |
| required_todos | ["depth-probe","optimized-prompt","execution","audit"] |
Transform rough questions into structured prompts, budget execution effort across phases, then execute against the plan.
| Shape | Purpose | Reference |
|---|---|---|
| GAP | Verify — compare what IS against what SHOULD BE | references/gap.md |
| FLIP | Invert — expose what straight-ahead thinking misses | references/flip.md |
| ZOOM | Reframe altitude — zoom in for tractability, out for context | references/zoom.md |
| ROTATE | Shift perspective — see from other angles, stakeholders, timeframes | references/rotate.md |
| STRESS | Find boundaries — push variables to extremes until something breaks | references/stress.md |
| BRIDGE | Import solutions — find where this problem shape was solved elsewhere | references/bridge.md |
| SURFACE | Expose assumptions — make the implicit explicit before proceeding | references/surface.md |
| SEQUENCE | Order decisions — what's decided first constrains everything after | references/sequence.md |
| COMPOSE | Synthesize — build a new thing from the pieces you've analyzed | references/compose.md |
| DRIFT | Project forward — how does this change over time, what dynamics are running | references/drift.md |
| RANK | Force-rank by impact — prioritize, don't just list | references/rank.md |
| Lens | Probe | Reference |
|---|---|---|
#orwell | "Say it in half the words" | references/lens-orwell.md |
#clausewitz | "Where does this break on contact?" | references/lens-clausewitz.md |
#miles | "What would you NOT include?" | references/lens-miles.md |
#popper | "What would disprove this?" | references/lens-popper.md |
#socrates | "Does that contradict what you said?" | references/lens-socrates.md |
#ive | "Does this feel inevitable or assembled?" | references/lens-ive.md |
#bezos | "What will 80-year-old you regret?" | references/lens-bezos.md |
#lamport | "What states haven't you specified?" | references/lens-lamport.md |
#feynman | "Can you explain this to a bright 12-year-old?" | references/lens-feynman.md |
#tufte | "What's the information density of this output?" | references/lens-tufte.md |
#taleb | "What's the downside you're not seeing?" | references/lens-taleb.md |
Multiple shapes: Order by: assumption-checking first, altitude-setting second, main analysis third, adversarial testing last.
| Signal | Mode |
|---|---|
--lite flag in args | Lite |
| Single question, straightforward | Lite |
| Multi-part, design, or complex task | Full |
| Ambiguous | Full (default) |
Step 1: Depth Probe → Read and follow steps/depth-probe.md
Step 2: Optimized Prompt → Read and follow steps/prompt.md
Step 3: Plan + Execute + Audit → Read and follow steps/full-mode.md
Step 2: Execute directly from depth_probe output. No optimized prompt artifact. No plan.
For each phase from your thinking tool selection:
If optimized prompt has >4 phases, read and follow references/delegate.md. Classify each phase RETAIN or DELEGATE.
| Skill | Use For |
|---|---|
optimizing-prompts | Optimize without executing |
ensemble-reasoning | When depth_probe reveals fundamental method flaw |
Surfaces interpretations before answering. Use when user appends #? or when question has multiple valid readings. Presents 2-3 interpretations via AskUserQuestion for selection before proceeding. For deep investigation, use mode-think.
Side-by-side comparison of options. Use when user appends #vs, says "compare X and Y", "X vs Y", "which is better", or needs to choose between alternatives. For honest yes/no verdicts, use mode-judge instead.
Save comprehensive context to a handoff file for future sessions. Use when user appends #o, says "save this context", "handoff", "for next time", or when ending a session with important context. Creates a persistent record.
Gives honest evaluation without cheerleading. Use when user appends #j, asks "good idea?", "should we", "worth it?", or needs unbiased assessment. No validation theater. For challenging approaches specifically, use mode-challenge instead.
Triple-checks proposals and claims before execution. Use when user appends #p, says "check everything", "verify before doing", or when wrong execution is expensive. Three passes: facts, completeness, consequences. For post-hoc evidence proving, use mode-verify instead.
Ruthless clarification context. Use when #qh prefix detected. Interrogates assumptions and forces explicit answers before proceeding. For lighter touch, use mode-question-light.
