grimoire-friction-management

name	grimoire-friction-management
description	Gestion active de la friction conversationnelle via QEC et budget de questions. Use when: too many questions, batching, friction budget, reduce interruptions, smooth execution, autonomous flow. Converts scattered clarifications into minimal high-value question batches.

Friction Management

Reduce conversational friction while preserving correctness.

When to Use

• User says there are too many questions or interruptions
• Work requires many clarifications across files/domains
• A long implementation risks stalling on micro-decisions
• Session momentum is dropping due to context switching

Friction Signals

Track these signals continuously:

• Number of clarification prompts per phase
• Repeated asks for information already available in workspace
• Latency caused by avoidable confirmation loops
• User prompts indicating impatience or flow break

Procedure

Step 1 — Build a Question Backlog

Before asking anything, collect all pending uncertainties and classify them:

• Blocking: cannot proceed safely without user input
• Non-blocking: can proceed using defaults and report assumptions
• Nice-to-have: defer unless user asks

Step 2 — Apply a Friction Budget

Use a strict budget per phase:

• Default budget: one batched clarification request
• If budget exceeded: choose safe defaults, continue, and log assumptions
• Escalate only for high-risk blocking decisions

Step 3 — Batch via QEC

When questions are needed, ask in one compact batch:

• Group by decision area
• Offer recommended default for each item
• Allow quick one-line user response

Step 4 — Report Assumptions Explicitly

For non-blocking uncertainties, proceed and surface assumptions in output:

## Assumptions Applied

- Decision area: ...
- Assumption used: ...
- Why safe: ...
- How to override: ...

Guardrails

• Never trade correctness for silence on high-risk operations
• Never ask one question at a time when batching is possible
• If uncertainty is high and impact is high, escalate immediately

Outcome Target

• Fewer interruptions
• Faster forward progress
• Clear assumption traceability