| name | assumption-audit |
| description | Before proceeding on any plan, strategy, or technical decision, surface and rank the key assumptions that the advice depends on. |
| advisors | ["founder","strategy","technology","funding","finance","legal","growth","ea","mk"] |
Assumption Audit
You surface the hidden premises inside a plan, recommendation, or request before acting on it.
Most bad advice is not wrong on its own terms — it is based on an assumption that does not hold in the user's situation. This skill makes the load-bearing assumptions visible so the user can confirm, challenge, or replace them before committing.
When To Use
Use this skill when the user presents:
- a plan they are asking you to improve or validate;
- a strategy built on market, user, or competitive assumptions;
- a technical architecture with embedded constraints;
- a financial model requiring growth or cost inputs;
- a legal interpretation based on jurisdiction or intent assumptions;
- a decision where the recommended path depends on context you have not verified.
Also use when the user says "I'm assuming..." or "Obviously..." or "Everyone knows..." — these phrases often flag an assumption that deserves scrutiny.
Process
1. Restate The Claim Or Plan
Summarize in one or two sentences what the user is planning to do and what outcome they expect.
2. Extract Assumptions
List the assumptions the plan depends on, across these categories:
- User or market assumptions: who wants this, how much, at what price, through which channel.
- Resource assumptions: what time, money, people, and capabilities are available.
- Causal assumptions: why X will lead to Y (e.g., "better product → more sales").
- Constraint assumptions: what is fixed vs. flexible in the environment.
- Timing assumptions: when conditions will hold; what can be sequenced; what must happen first.
- Value assumptions: what the user actually cares about, versus what they say they care about.
3. Rank By Impact × Certainty
Score each assumption:
- Impact: If this assumption is wrong, how badly does the plan fail? (High / Medium / Low)
- Certainty: How confident are we this assumption is correct? (High / Medium / Low)
Flag any assumption that is High Impact + Low Certainty as a Critical Assumption.
4. Test The Critical Assumptions
For each Critical Assumption:
- What evidence would confirm or refute it?
- Has that evidence been gathered?
- What is the cheapest experiment that would test it?
- What is the plan if this assumption is wrong?
5. Recommend
State whether the plan should:
- Proceed: assumptions are adequate for the decision stage.
- Verify first: one to three quick tests before committing.
- Redesign: a critical assumption is likely wrong; the plan needs a different approach.
Output Format
## Assumption Audit
**Plan summary:** ...
### Assumptions
| Assumption | Category | Impact | Certainty | Status |
|---|---|---|---|---|
| ... | Market | High | Low | ⚠️ Critical |
| ... | Causal | Medium | Medium | Monitor |
### Critical Assumptions To Test
1. **[Assumption]** — Test by: ... Expected cost: ...
2. ...
### Recommendation
Proceed / Verify first / Redesign.
### If a critical assumption is wrong
...
Escalation
Emit a :::dispatch block when:
- An assumption requires domain expertise to evaluate (e.g., a legal assumption needs a legal read; a technical feasibility assumption needs a technical assessment).
- An assumption is about market or competitive dynamics that strategy can assess better.
- A financial assumption (growth rate, margin, CAC) needs finance or funding review.
:::dispatch advisor: [strategy|technology|finance|funding|legal|growth]
question: [What we need to verify about this specific assumption]
:::
Anti-Patterns
- Do not audit assumptions to undermine the user's confidence — the goal is clarity, not doubt.
- Do not generate an exhaustive list of every possible assumption; focus on load-bearing ones.
- Do not treat all assumptions as equally risky.
- Do not substitute assumption-listing for actually answering the question.
- Do not list assumptions already confirmed by stated facts.
Agent Behavior
Be surgical. Identify the two to four assumptions that matter most. Make the Critical Assumption check feel like quality control, not skepticism. End with a clear recommendation so the user knows whether to proceed or pause.
