Decision Quality Audit: AI Skill for Product Management
Decision Quality Audit is an AI skill that reviews product decisions on the reasoning, not the outcome. It names the bias driving a call and rates each finding by decision damage, with a concrete fix.
npx skills add Uxcel-Lab/product-skills --skill pm-decision-quality-audit
What this skill does
Judges product decisions on the reasoning available at the time, not on how the outcome happened to land.
- Separates process quality from outcome luck, so lucky wins stop teaching bad habits.
- Checks deliberation matched reversibility: fast on two-way doors, careful on one-way doors.
- Requires calibrated confidence levels and weighs the cost of delay.
- Names the bias driving the call and pairs it with a structural countermeasure.
- Breaks the sunk-cost trap: starting from zero today, would you still choose this?
- Checks high-stakes bets ran a pre-mortem and recorded rejected alternatives.
- Verifies the reasoning is documented before the outcome is known.
- Rates every finding by decision damage and ends with the top three fixes.
When to use it
Use it when you need an AI assistant to:
- audit a build, buy, sunset, re-platform, or pricing call;
- review a decision log or retrospective without hindsight bias;
- pressure-test a high-stakes bet before committing;
- name the bias quietly steering a choice;
- test whether sunk cost is keeping a weak bet alive;
- set up reasoning a team can review honestly later.
What this audit checks
| Category | What it catches |
|---|---|
| Resulting | Decisions graded by what happened next instead of the reasoning available at the time |
| Reversibility fit | Committee ceremony on two-way doors, casual calls on one-way doors |
| Calibrated confidence | Binary verdicts, analysis paralysis, ignored cost of delay |
| Bias | Confirmation, anchoring, framing, false consensus, hindsight |
| Sunk cost | Past investment used as a reason to continue |
| Pre-mortem & trade-offs | High-stakes bets with no failure modes named and no rejected alternatives recorded |
| Documentation | No written rationale, confidence level, or scheduled review |
How the skill works
1
Frame the decision
The skill confirms the decision, its reversibility, and whether it reviews before commitment or in retrospect.
2
Check the reasoning
It separates process from outcome, names active biases, and tests confidence, trade-offs, and pre-mortems.
3
Deliver ranked findings
It rates each finding by decision damage and lists the top three fixes with questions to answer.
Try it with
Related skills
1.0.0
MIT
Product Management
Audits
