TLDR
- A compact add-on you paste under any base prompt to surface blindspots, trade-offs, risks, biases, compliance issues, and fast validation steps.
- Produces a prioritized questions list, assumption register, risk and kill switch checklist, and a tiny validation plan with next actions.
- Works in ChatGPT, Claude, Gemini, and Copilot without bloating the main answer.
Introduction
Most answers skip the messy middle: unknowns, risks, and trade-offs. Blindspot Booster Pro v2.1 is a drop-in add-on that forces a disciplined blindspot pass after your base answer. It flags assumptions, proposes low-cost tests, and returns decision-ready next steps. Paste it after your prompt, adjust a few brackets if you like, run it, then act on the three concrete actions at the end.
The Custom Prompt
Master Add-On: Blindspot Booster Pro v2.1
[NAME] Blindspot Booster Pro — Unknown-Unknowns Add-On (v2.1)
[ROLE] Act as a rigorous blindspot detector, red-team partner, and strategist.
[APPLICATION RULES]
- First produce or assume the base answer (internally). Do NOT repeat it.
- Then run this blindspot pass. Add net-new insight only.
- Do not reveal internal chain-of-thought; give concise conclusions with brief justifications.
[SCOPE DIALS]
- Mode: [Decision-ready | Exploratory] (if Decision-ready, demand stricter evidence)
- Decision Type: [Two-way door (reversible) | One-way door (hard to reverse)]
- Reversibility: [low/medium/high] — prefer reversible options under high uncertainty
- Uncertainty Budget: [where rough is allowed vs. must be precise]
- Time Horizons: [now, +1 week, +3 months] — tag items by horizon and likely to change triggers
[PARAMETERS]
- Depth: [2/3] (1 quick, 2 thorough, 3 exhaustive)
- Risk Appetite: (low/medium/high)
- Evidence Standard: [separate facts vs. assumptions; cite if known; label “Speculative”]
- Question Quota: [7] (max clarifiers at end, only if essential)
- Cost/Time Budget: [$0–$X, ≤Y hours] for proposed validation
- Output Mode: [bullets + json] (json-only/bullets-only)
- Safety/Compliance Toggles: [privacy, security, accessibility, ethics, localization, sustainability, IP/licensing] = [on/off per item]
- Bias/Harm Check: [on] (scans for cognitive bias, harm, misuse; propose mitigations)
[TASKS] Provide the sections below:
1) Unasked but Important Questions — P1–P3 prioritized, 1-line why, tag horizon.
2) Assumption Register — each with owner, review date, criticality, validation method.
3) Trade-off Explorer — competing objectives; Pareto note; recommended compromise and rationale.
4) Baselines and Comparators — prior art, current best known approach, benchmark gap.
5) Risks and Failure Modes — likelihood × impact × detectability; early signals; kill-switch criteria.
6) Counterfactuals and Reversibility — what if the opposite is true; cheapest falsification test; rollback plan.
7) Lenses and Stakeholders — regulator or ethics, operator or maintainer, adversary or abuser, accessibility or inclusion, localization, sustainability, IP or licensing.
8) Data Gaps and Acquisition Plan — missing facts, sources, cost or latency, privacy or security constraints.
9) Metrics and Guardrails — success metrics, threshold ranges, do-not-exceed caps, review cadence.
10) Validation Plan — smallest tests this week; expected signal; stop or go criteria; cost or time fit.
11) Decision Checklist — must-answer items, approvals or DRI, documents or artifacts to assemble.
12) Next 3 Actions — tiny, concrete, immediately doable steps.
[CONSTRAINTS]
- Be specific and practical; no hand-waving.
- Label uncertain items “Speculative.” Avoid confident guesses.
- Respect Safety or Compliance toggles and Cost or Time Budget.
- Ask ≤ [Question Quota] clarifiers only if essential, and say why each matters.
[BIAS/HARM CHECK]
- Scan for confirmation bias, sunk cost, scope neglect, survivorship bias.
- Identify potential harms or misuse or abuse; propose fences or mitigations.
[SELF-SCORE AND IMPROVE]
- Rate this blindspot pass 1–5 on: Novelty, Specificity, Prioritization, Rigor, Actionability.
- Suggest one specific improvement and incorporate it briefly.
[OUTPUT FORMAT]
- Sections as bullets PLUS a compact JSON object:
{
"unasked_questions": [{"priority":"P1|P2|P3","q":"...", "why":"...", "horizon":"now|+1w|+3m"}],
"assumptions": [{"assumption":"...", "owner":"...", "review_date":"YYYY-MM-DD", "criticality":"high|med|low", "validation":"..."}],
"tradeoffs": [{"objective_a":"...", "objective_b":"...", "current_position":"...", "alternatives":["..."], "rationale":"..."}],
"baselines": {"prior_art_refs":["..."], "current_best":"...", "benchmark_gap":"..."},
"risks": [{"mode":"...", "likelihood":"low|med|high", "impact":"low|med|high", "detectability":"low|med|high", "early_signal":"...", "mitigation":"...", "kill_switch":"..."}],
"counterfactuals": [{"statement":"...", "test":"...", "rollback_plan":"..."}],
"lenses": {"regulator":"...", "operator":"...", "adversary":"...", "accessibility":"...", "localization":"...", "sustainability":"...", "ip_licensing":"..."},
"data_plan": {"gaps":["..."], "sources":["..."], "cost":"...", "latency":"...", "privacy_constraints":"..."},
"metrics_guardrails": {"metrics":[{"name":"...", "target":"..."}], "guardrails":[{"name":"...", "cap_or_floor":"..."}], "review_cadence":"..."},
"validation_plan": [{"experiment":"...", "expected_signal":"...", "stop_go":"...", "cost_time_fit":"..."}],
"decision_checklist": ["..."],
"next_actions": ["...", "...", "..."],
"clarifiers": [{"q":"...", "why":"..."}],
"self_score": {"novelty":1,"specificity":1,"prioritization":1,"rigor":1,"actionability":1},
"improvement_suggested": "..."
}
Variants (pick one if you want it shorter)
A) Concise Drop-In Lite
Depth 1. Output only:
• Top 5 Unasked Qs (P1–P3, 1-line why, horizon)
• 5 Key Assumptions (mark Critical)
• 3 Risks (plus early signal and mitigation)
• 2 Trade-offs or Alternative Framings
• 3 Next Actions
Label uncertainties “Speculative.” No repetition of base answer.
B) Red-Team Pulse (adversarial)
In ≤8 bullets: cheapest attack; likely abuse paths; data or privacy weak points; single riskiest assumption plus falsification test; rollback plan (48-hour pivot); top kill-switch trigger.
C) Regulated-Domain Mode (compliance-first)
Toggles ON: privacy, security, accessibility, ethics, IP or licensing.
Include: applicable obligations (for example HIPAA or GDPR or SOC2 or WCAG) — “Advisory, not legal advice”; data handling map; DPIA or threat model sketch; access controls; audit artifacts; go or no-go checklist.
D) Programmatic JSON-Only
Return JSON only per the schema in OUTPUT FORMAT. No prose.
Acceptance criteria
• Adds new, material insights or questions not in the base answer.
• Separates facts vs. assumptions; labels speculation.
• Prioritizes P1s; references reversibility or decision type.
• Proposes at least one cheap, time-bounded validation step.
• Honors Cost or Time Budget and Safety or Compliance toggles.
• Provides a self-score and one improvement.
Evaluation rubric 1–5
Novelty • Specificity • Prioritization • Rigor • Actionability • Safety or Compliance Fit
Simulation preview
You will get a compact, prioritized set of unasked questions, hidden assumptions with owners and dates, trade-offs with a recommended compromise, baseline comparisons, top risks with early signals and kill-switches, a tiny validation plan that fits your budget, and three next actions, plus a brief self-score and one improvement.
Platform adapter notes
• ChatGPT: works as-is.
• Claude: loves longer context, keep Constraints explicit.
• Copilot: use Variant A or D; say “Use bullet points.”
• Gemini: add “think step by step, then summarize” if you want extra scaffolding.
What This Add-On Does
- Surfaces blindspots without repeating your answer: runs a net-new pass to expose risks, trade-offs, and assumptions.
- Adds decision hygiene: provides kill switches, reversibility notes, and stakeholder lenses that reduce regret.
- Keeps it lightweight: outputs a short bullet set plus a compact JSON that you can drop into Notion or Sheets.
Step by Step Usage
- Write your base prompt and run it.
- Paste the Blindspot Booster Pro block under it.
- Optionally set Mode, Depth, Risk Appetite, and Cost or Time Budget.
- Run, then act on the Next 3 Actions and the Validation Plan this week.
- Save the JSON to your tracker and schedule the next review date.
FAQ
Will this slow us down
No. Use Variant A when you need one screen. It still returns questions, risks, and next actions.
How does it handle compliance
Switch on the relevant Safety or Compliance toggles. The add-on lists obligations, data handling maps, and a go or no-go checklist.
What if we lack data
Uncertain items are labeled Speculative and routed to the Data Gaps and Acquisition Plan with low-cost sources.
Conclusion
Blindspot Booster Pro v2.1 turns quick answers into safer decisions. It exposes unknowns, separates facts from assumptions, and gives you a tiny, budget-fit validation plan with owners and dates. Use it as a default suffix for plans, policies, and designs. It will help your work get cited for its rigor and clarity while staying lightweight and practical.
Field Drill Walkthrough
Scenario: Internal policy update with privacy impact
- Mode: Decision-ready, Depth 2, toggles privacy, security, accessibility on.
- Output: P1 questions on data retention and access controls, a kill switch tied to audit findings, and a two-step validation plan: policy dry run with five users, then targeted DPIA notes.
- Next 3 Actions: confirm owners and review dates, run the dry run, file the decision in the log.
References and Links
- HBR: Performing a Project Premortem: https://hbr.org/2007/09/performing-a-project-premortem
- NIST Risk Management Framework: https://csrc.nist.gov/publications/detail/sp/800-37/rev-2/final
- UK ICO: Data Protection Impact Assessment guidance: https://ico.org.uk/for-organisations/guide-to-data-protection/impact-assessments-dpia/
- Cognitive biases overview: https://en.wikipedia.org/wiki/Cognitive_bias
Inversion Engine: Opposite-Plan + Option Value + Governance (v3)
TLDR Introduction Most teams ask how to make Plan A work. Fewer ask what happens if we choose the exact opposite. Inversion...