メニュー
Lean, evidence-backed project Q&A for advanced coding models. Use when the user wants to understand repository facts, architecture, status, impact, or workflow choices without changing files.
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
Advanced post-Review human acceptance workflow. Use after fresh approved Review closeout to restore a contextless human's understanding, guide one real product step at a time, and persist an explicit pass, fail, or blocked verdict.
Read-only cross-artifact consistency analysis for advanced coding models. Use when tasks exist and the specification, plan, and task boundaries need an independent gate before or during execution.
Existing-spec repair workflow for advanced coding models. Use when a specification package has planning-critical ambiguity, weak acceptance, contradictions, or new constraints that should be absorbed without recreating the feature.
Focused pre-plan feasibility research for advanced coding models. Use when a planning-ready specification still lacks a credible implementation chain, external evidence, or a disposable proof.
Lean design-system workflow for advanced coding models. Use for a new product UI, redesign, rebrand, shared visual language, or an audit/update of the root DESIGN.md contract.
Resumable product and technical discussion for advanced coding models. Use when a rough idea needs durable decisions and options before formal specification.
| name | spx-ask |
| description | Lean, evidence-backed project Q&A for advanced coding models. Use when the user wants to understand repository facts, architecture, status, impact, or workflow choices without changing files. |
Read references/project-learning.md and apply its consume-only policy.
Read references/project-cognition.md, using cognition intent ask.
Read references/evidence-contract.md for architecture, status, or impact
claims. Route a request to interpret generated workflow artifacts or lane state
to $spx-explain.
Answer the question from the smallest sufficient live repository evidence set. Lead with the conclusion, cite concrete paths, and distinguish verified facts, evidence-backed inference, and remaining unknowns. Project cognition selects where to look; the live repository proves the answer.
Remain read-only. Do not create workflow state, handoffs, reports, memory, or planning artifacts. Avoid commands that mutate dependencies, caches, generated assets, or application state. Recommend another SPX skill only when the user is asking for follow-on action; do not switch workflows automatically.
