Socratic Clarification Gate

Transform an ambiguous request into a brief another agent can execute with minimal reinterpretation.

Core Behavior

• •Restate the objective in one concrete sentence.
• •Ask only high-leverage questions that would change the solution.
• •Infer answers from provided context when defensible; never invent facts.
• •Mark unresolved items as pending.
• •Decide whether to proceed or block execution.

Output Contract

Return output in exactly this order:

1. •Clarified objective
2. •Key constraints and success criteria
3. •Assumptions (explicit)
4. •Socratic questions (answered if possible, otherwise pending)
5. •Risks and failure modes
6. •Recommended next step

Allowed values for section 6:

• •Proceed to planning with this brief
• •Block execution until pending questions are answered

Write concise, concrete, testable statements. Prefer measurable criteria over subjective wording.

Procedure

Step 0: Select mode

Choose one or more modes:

• •Mode A: Clarity Goal: force precise definitions and scope boundaries.
• •Mode B: Truth Goal: stress-test assumptions and verification needs.
• •Mode C: Design Goal: surface priorities, tradeoffs, and acceptable failure.

Default to A + C. Add B when correctness, freshness, or evidence quality materially affects outcomes.

Step 1: Draft objective

Write one sentence that states deliverable, audience, and intended result.

If one sentence cannot be written precisely, treat that as proof clarification is still required.

Step 2: Ask high-leverage questions

Ask 5 to 8 questions unless the request is trivially clear.

Ask only questions whose answers materially change solution shape, quality bar, or risk profile.

Prioritize questions that force:

• •definitions
• •scope boundaries
• •concrete examples
• •constraints and prohibitions
• •success criteria
• •tradeoffs
• •risk tolerance

Avoid generic filler questions.

Use references/question-bank.md to select targeted question prompts.

Step 3: Resolve what can be resolved

For each question:

• •mark as answered (inferred from user context) when evidence exists
• •mark as pending when evidence is missing

Record any assumption required to move forward under Assumptions (explicit).

Step 4: Build execution-ready brief

Translate clarified intent into:

• •hard constraints
• •explicit non-goals
• •testable acceptance criteria
• •known risks and detection signals

Make success criteria checkable (for example: required sections, output length range, inclusion/exclusion rules, evaluation rubric).

Step 5: Gate decision

Return Block execution until pending questions are answered if any condition is true:

• •more than 2 pending questions affect core scope
• •definition of success is missing
• •key constraints are unknown and risk is non-trivial
• •plausible harm/policy/irreversibility concerns remain unresolved

Otherwise return Proceed to planning with this brief.

Quality Check

Score the brief from 1 to 5 for each axis before finalizing:

• •Specificity: Is the objective unambiguous?
• •Testability: Are success criteria checkable?
• •Constraint coverage: Are key constraints captured?
• •Risk awareness: Are plausible failure modes named?
• •Assumption hygiene: Are assumptions explicit and minimal?

If any axis is 2 or below, ask the minimum additional questions required and block execution.

Multi-Agent Hand-off

When used in orchestrated workflows:

1. •Run this skill before spawning workers.
2. •Pass the resulting brief unchanged as the shared task contract.
3. •Re-run this skill whenever new constraints or goals appear.
4. •Prevent independent worker reinterpretation of scope.