Scope Negotiator

Shrink messy work into a useful slice. This skill is for scope control, not for deep requirements interviewing or full implementation planning.

Workflow

1. Name the broad request.

   • Restate the user's desired outcome in one sentence.
   • Identify why the request could expand: unclear audience, too many artifacts, too many repos, quality bar, dependencies, or hidden decisions.

2. Separate scope layers.

   • Core outcome: the smallest result that creates real value.
   • Required support: checks, docs, migrations, tests, or artifacts needed for trust.
   • Optional expansion: polish, automation, extra formats, broad refactors, future-proofing.
   • Explicit non-goals: tempting work that should stay out for now.

3. Propose 2-3 scope options when useful.

   • Minimal safe slice: fastest useful outcome with basic verification.
   • Standard slice: balanced outcome likely matching the user's intent.
   • Expanded slice: only when the extra work changes long-term leverage.

4. Recommend one slice.

   • Choose the smallest slice that satisfies the user's apparent goal and risk level.
   • If the user asked for execution and the recommended slice is safe, proceed after stating the chosen scope.
   • Ask one concise question only when choosing the wrong slice would waste substantial work or violate a boundary.

5. Preserve future work.

   • List deferred items so they are not forgotten.
   • Route to $define-goal when the user wants a goal-backed objective.
   • Route to $codex-consensus-plan when the chosen scope still needs a decision-complete implementation plan.
   • Route to $codex-completion-loop when the scope is clear and should be executed through verification.

Output Shape

Use this structure:

• Recommended scope
• Why this slice
• In scope
• Out of scope
• Deferred
• Acceptance criteria
• Next action

When comparing options, use:

Option | What it includes | Cost | Risk | When to choose

Scope Rules

• Prefer current-turn value over theoretical completeness.
• Do not force a large plan when a small reversible edit or answer solves the need.
• Do not silently drop important work; defer it explicitly.
• Keep non-goals visible when they prevent scope creep.

Stop Conditions

• Stop negotiating once the current slice is clear enough to execute.
• Do not ask broad preference questions when repo context or user history provides a safe default.
• Do not use this skill to avoid doing clear implementation work.