Do you want to guess how something will turn out?

I. Q&A and Setup

1. Create a subdirectory for this workflow's output before doing anything else.
2. Conduct brief web research to understand the current context of the user's topic before asking questions.
3. Conduct a Q&A to establish the current situation and define the analytical boundaries. 3a. Read and apply the conversational framework in guess_how_it_will_turn_out_brainstorming skill. 3b. Walk the user through all four conversational rounds defined in the SKILL file. Do NOT compress or skip rounds. 3c. Once the user has chosen the final binary or mutually-exclusive forecasting question in Round 4, generate a setup file named 00_initial_setup_and_options.md: 3ci. A Background Facts section capturing the timeline, actors, and constraints established during the brainstorm. 3cii. An Alternative Questions section logging the unpicked questions from the menu for future reference. 3ciii. An Official Anchor section declaring the final, approved forecasting question. This exact question anchors the remainder of the analysis.

II. Global Execution Directives

1. Whenever you write content to a file, also show the output on the chat screen.
2. When switching engines, use the respective handover skills (switch-to-claude-opus-4.6-thinking workflow or switch-to-gemini-3.1-pro-high workflow).
3. This workflow has exactly two permitted pauses: Step I-2 (the Q&A interview) and Step VI-2 (the optional external audit prompt). Outside of those two moments, execute all steps continuously without stopping or asking questions. Print clipboard warnings and interim notices inline — never pause to ask for confirmation.
4. Analytical Integrity: When constructing affirmative/negative cases and simulating histograms, read and strictly apply analytical_integrity skill.

III. The Simulation & Memos

1. Research: Thoroughly research the Official Anchor question using web search and any other available tools. Think carefully about what information you need to predict the outcome and find it. If you can't find it online, ask the user. Make sure it's current to the minute. Be thorough — capture every relevant influence (legal, social, political, economic, or any other field). Save as 01_research.md.
2. The Affirmative Case: Find every possible reason why the Official Anchor outcome WILL happen. In addition to your own reasoning, conduct independent web research to identify the strongest real-world arguments. Be thorough — in a prior run you missed a widely known reason. Save as 02_why_it_will_happen.md.
3. The Negative Case: Find every possible reason why the Official Anchor outcome WON'T happen. Same standard of thoroughness and independent research as above. Save as 03_why_it_wont_happen.md.
4. 1,000-World Synthesis: Synthesize the affirmative and negative memos to estimate what will actually happen: 4a. Create 1,000 identical starting "worlds" based on all facts in the setup file. Run them forward and see whether the anchor event happens or not in each run. 4b. Categorize outcomes into a "yes" or "no" bucket, then split further into descriptive subcategories. You choose the subcategories that make sense. N can be as few as 2 or as many as 20 (e.g., 8 yes buckets and 12 no buckets). 4c. Label each bucket with [YES] or [NO] and state its percentage of runs. 4d. After the histogram, state the total: "X% YES / Y% NO." 4e. Below the histogram, create a bulleted list describing each bucket. Save as 04_synthesis_and_histogram.md.
5. Empirical Sanity Check: Research empirical data, historical facts, studies, or any real-world evidence to see if the histogram is significantly off base. Go as far back in history as appropriate and identify anything even loosely relevant (weight it proportionally to how analogous it actually is). Compare findings to the predictions. Save as 05_empirical_sanity_check.md.

IV. Peer Review Phase

1. Complete all file writes first, then switch engines: 1a. Use the switch-to-claude-opus-4.6-thinking workflow skill. 1b. The new engine reviews ALL four memos (02, 03, 04, 05), one at a time. Verify arguments, challenge assumptions, and save revised memos with a _reviewed suffix (e.g., 02_why_it_will_happen_reviewed.md). Do NOT pause for user review. Proceed directly to step 2.
2. Switch back: 2a. Use the switch-to-gemini-3.1-pro-high workflow skill.

V. Final Summarization

1. Create a concise bullet-point summary structured as follows: 1a. The Core Question: Define the exact anchor question being predicted. State what "YES" and "NO" mean in context so the histogram is unambiguous. 1b. Why it will happen: Succinct bullet points from the reviewed affirmative memo. 1c. Why it won't happen: Succinct bullet points from the reviewed negative memo. 1d. The histogram: The final revised distribution, ordered most-to-least likely, each bucket labeled [YES] or [NO], with a total percentage rollup. 1e. Bucket explanations: A multi-sentence explanation of each bucket, formatted as bullet points (not paragraphs). Each bucket gets a bold heading with its label and percentage, followed by bullets explaining: (a) the specific mechanism or chain of events, (b) the key evidence or historical analogues, and (c) why it received its probability weighting. Do NOT use vague one-liners like "Japan 1941 logic" — explain what actually happens in that scenario and why. 1f. Empirical validation: Brief summary of the key data points from the empirical memo that anchored or adjusted the prediction. 1g. Save as 06_summary.md.

VI. External AI Audit

1. The Compilation Package: Create a compilation file designed to be copy-pasted into Google Deep Think (or another external reasoning AI) for a hostile peer review. Include the final versions of all reviewed memos in order. At the top, explain what you did and request a review. In that prompt, you MUST command the external AI to generate its own alternative histogram: "You must provide your own histogram of possible outcomes. You do not have to use our buckets — you can create your own — but you must provide a full percentage breakdown of what you think will happen." Save as 07_external_audit_package.md.
2. External Review (Optional): PAUSE HERE — Ask the user: "Do you want to run the audit package through Gemini Deep Think, Deep Research, both, or skip?" Wait for their answer. Once they answer, proceed continuously without further pauses. If yes: 2a. Use 07_external_audit_package.md as the source material, instructing critical review of arguments, probabilities, and empirical grounding. 2b. Run the chosen tool(s) via Gemini Deep Think or Gemini Deep Research by using the gemini_google_com_browsing skill, monitoring the browser tab automatically until complete. Move directly to step 2c — do not stop to discuss results with the user. 2c. Review the external audit outputs and act as an Adjudicator. Revise the summary memo: 2ci. Create _proposed.md by surgically editing the existing summary (do NOT rewrite from scratch). 2cii. Delta Analysis: At the top of the V2 summary, prepend an "Adjudication Stance & Delta Analysis" section. Declare one of three stances: [Full Incorporation] (audit proved baseline fatally flawed; histogram radically shifted), [Partial Synthesis] (audit had valid points but extrapolated too far; histogram moderately shifted), or [Hard Rejection] (audit was flawed; histogram unchanged). Explain exactly why the logic was accepted or rejected and what specific factors changed. 2ciii. Save as 06_summary_v2.md. 2d. Preservation rule: NEVER overwrite the original files. The v1 and v2 results must coexist so the user can compare baseline against the audit's impact. 2e. Clean up scratch files (_reference.md, _proposed.md) after final versions are confirmed.

• 00_initial_setup_and_options.md — Background facts, alternative questions, and the official anchor forecasting question.

  • Example: 001_prediction_will_x_happen\​00_initial_setup_and_options.md

• 01_research.md — Thorough research findings on all factors relevant to the prediction.

  • Example: 001_prediction_will_x_happen\​01_research.md

• 02_why_it_will_happen.md — The affirmative case: every reason the outcome WILL happen.

  • Example: 001_prediction_will_x_happen\​02_why_it_will_happen.md

• 03_why_it_wont_happen.md — The negative case: every reason the outcome WON'T happen.

  • Example: 001_prediction_will_x_happen\​03_why_it_wont_happen.md

• 04_synthesis_and_histogram.md — The 1,000-world Monte Carlo simulation with percentage breakdown by outcome bucket.

  • Example: 001_prediction_will_x_happen\​04_synthesis_and_histogram.md

• 05_empirical_sanity_check.md — Historical precedents and empirical data used to validate the prediction.

  • Example: 001_prediction_will_x_happen\​05_empirical_sanity_check.md

• 06_summary.md — The final bullet-point summary with the histogram, bucket explanations, and empirical validation.

  • Example: 001_prediction_will_x_happen\​06_summary.md

• 07_external_audit_package.md — A compilation file designed to be pasted into an external AI for hostile peer review.

  • Example: 001_prediction_will_x_happen\​07_external_audit_package.md