Quizme Mode
Mission
Create a conversation-local clarification gate that prevents substantive execution until the user and agent have a fully aligned task contract.
Trigger Rule
Use this skill when:
- the user writes
--quizme - quizme mode was enabled earlier in the active conversation and has not been toggled off
Trigger parsing is exact:
--quizmetoggles quizme mode on when off--quizmetoggles quizme mode off when already on- mode persists throughout the active conversation until toggled off
- only arguments placed directly after
--quizmeare considered - supported arguments are
--mc,--one-at-a-time,--confirm, and--record - arguments may appear in any order
- duplicate arguments are harmless
- unsupported arguments make the command invalid; briefly report them, leave state unchanged, and do not consume the message
- toggling quizme mode off clears every option
- while mode is active, plain
--quizmetoggles it off;--quizmewith supported arguments keeps it on and replaces active options
Multiple-Choice Preference
When invoked as --quizme --mc:
- enable quizme mode if it is off
- prefer multiple-choice clarification questions
- ask short-form questions only when multiple-choice options would distort the answer
- include an optional short-form choice when listed options may be incomplete
Optional Behaviors
--one-at-a-time:
- ask exactly one adaptive clarification question per round
- use each response to shape the next question
- do not reduce clarification depth
--confirm:
- show the finalized task contract after clarification
- wait for explicit user approval before substantive execution
--record:
- implies
--confirm - persist the approved contract in governance evidence or the most suitable task artifact when one exists
- do not create a durable artifact for answer-only tasks unless the user asks
Automatically require confirmation for destructive, public, production, payment, authentication, or irreversible tasks even when --confirm was not supplied.
Clarification Gate
Before substantive execution, keep questioning until no material doubt remains about:
- goal
- in-scope work
- out-of-scope work
- constraints
- acceptance criteria
- risk tolerance and side-effect boundaries
Questioning rules:
- ask one to three high-value questions per round unless
--one-at-a-timeis active - continue follow-up rounds while material uncertainty remains
- do not ask about facts safely discoverable from repository or environment inspection
- do not ask questions whose answers cannot change execution
- do not start implementation, mutation, or governed execution until the clarification gate passes
Interactive Clarification
Use the same interactive clarification flow as plan-mode questioning when the runtime exposes it.
- prefer the interactive clarification console for quizme questions
- use mutually exclusive multiple-choice options when
--mcis active - allow short-form input only when needed or as an optional additional choice
- if the interactive clarification console is unavailable, ask concise questions in normal conversation
Scope Boundary
This skill owns:
- toggle state
- immediate argument parsing
- clarification depth
- pre-execution blocking behavior
Use Requirement Clarifier for the resulting implementation contract. Use Skill Governance to record active quizme state for governed tasks.
Output Contract
Before execution begins, confirm:
- quizme mode state (
onoroff) - multiple-choice preference (
onoroff) - one-at-a-time preference (
onoroff) - confirmation requirement (
onoroff) - recording requirement (
onoroff) - clarified task contract
- remaining open questions (
nonebefore substantive execution)
Anti-Overuse Rules
Use when:
- the user explicitly invokes
--quizme - quizme mode remains active from an earlier turn
Do not use when:
- the user has not enabled quizme mode
- quizme mode was toggled off
Stop after:
- the task contract is fully clear
- substantive execution may begin for the current request
- quizme mode remains active for later requests until explicitly toggled off