Assumption Audit

Elicitation Entry Rule

BEFORE asking anything in this module:

1. Read elicitation-state.yaml (at docs/ets/projects/{project-slug}/state/ or knowledge/ template)
2. Find this module's entry in module_state
3. If status == "covered" → summarize what exists, ask if user wants to go deeper
4. If status == "in_progress" → continue from current_probe
5. If status == "not_started" → start with story probe (see .claude/skills/discovery/knowledge/story-probes.md)
6. After each question → update elicitation-state.yaml

PURPOSE

Separate what is known from what is merely believed.

This skill turns hidden uncertainty into explicit ASM-# records with status and validation paths. It is the main protection against false confidence in the ideation phase.

Assumption Status Options

Each ASM-# must have one of these statuses:

• resolved — Confirmed answer exists. Evidence documented.
• open — Conscious gap. Has owner + deadline for resolution.
• deferred — Conscious decision to defer. Has justification + target phase/milestone.
• assumed_default — Accepted during fatigue management (R10). Marked with revisit_required: true. Cannot be used for mandatory dimensions.
• unconscious_gap — Discovered via "I never thought about this" response. Was deepened with 2-3 follow-up questions before proceeding.

OUTPUT CONTRACT

For each ASM-#, capture:

• statement
• evidence available
• status: resolved | open | deferred | assumed_default | unconscious_gap
• risk if false
• validation needed
• owner / phase if known

AUDIT QUESTIONS

Ask one at a time:

• "What are we assuming about user behavior?"
• "What are we assuming about technical feasibility?"
• "What are we assuming about data quality or integrations?"
• "What are we assuming about approvals, policies, or dependencies?"
• "If this assumption is false, what breaks first?"

ANTI-PATTERNS

• Do not hide assumptions inside prose
• Do not mark something as resolved without evidence
• Do not leave a high-risk unknown without a validation path