دليل Fable Brain: حوّل ذكاء Fable 5 إلى تعليمات دائمة

You are the strongest model I currently have access to. I want to preserve the best parts of your judgment as reusable operating instructions for another capable model.

Write a standing-instructions document that the replacement model can run on every task I give it.

Important boundaries:
- Do not reveal private chain-of-thought or hidden internal reasoning.
- Instead, produce concise, executable procedures, checks, decision rules, and output standards.
- Address the entire document TO the replacement model, in second person, as commands it can execute.
- My instructions inside an individual task always override these standing instructions.

Cover these 10 areas, in this exact order:

1. Reading intent: how to identify what I actually need when my request is vague, messy, incomplete, or aimed at the wrong question. Include the exact trigger for asking one clarifying question instead of guessing.

2. Breaking problems down: how to split a difficult task into independently checkable parts, define dependencies, and choose the order of execution.

3. Effort placement: how to find the part where an error would cause the most harm, and allocate the most verification effort there instead of treating every part equally.

4. Verification: how to re-check every number, date, calculation, factual claim, source, link, and named entity before relying on it. Never trust a figure just because the sentence around it sounds fluent.

5. Known vs guessed: how to label inside the answer what is verified, what is a reasonable inference, what is an assumption, and what is unknown. Give exact wording for each level.

6. Self-attack: how to argue against your own conclusion before sending it, identify the strongest alternative explanation, and revise when the attack exposes a weakness.

7. Completeness: how to confirm that every part of a multi-part request was answered, every requested deliverable exists, and nothing was silently dropped.

8. Refusing to guess: the exact conditions where saying "I do not know" or asking for more information is better than producing a confident answer.

9. Delivery: how to give the direct answer first, the supporting explanation second, and risks or caveats last, using plain language and the user's requested level of detail.

10. Fake competence: list the 10 most common ways an AI produces answers that look correct but are not. For each one, include:
- the warning sign that exposes it
- the verification step
- the corrective action

Format rules for every area:
- Write each procedure as a trigger and an action: "When you see X, do Y."
- Every rule must be executable step by step. If a rule sounds like vague advice such as "be careful," rewrite it as a concrete action.
- Give one short worked example per area showing the procedure catching a real mistake.
- Name the failure each procedure prevents.
- Keep the rules reusable across research, writing, analysis, business work, and coding.

End the document with a FINAL SEND GATE: a short checklist the replacement must run before every answer. Include this rule: if any item fails, fix it and run the checklist again. Never send anyway.

Be exhaustive on substance and ruthless on length. Remove anything a strong model would already do without being told.

At the end:
1. Output the complete document in Markdown.
2. Give it the filename: fable-brain.md
3. If you run out of room, stop only at the end of a section and write: "Reply CONTINUE for the next section."

Use the standing instructions in this project.

Task: Review the following plan and identify the highest-risk assumption, verify every number, list missing information, and give me the direct recommendation first.

[PASTE YOUR PLAN HERE]