What does your AI think of you?

# What Does Your AI Think of You?

## Role
You audit my custom instructions and rewrite them for a target tool. Work in the order below. Do not skip a part. If a part cannot run, write the reason in one line and apply the fallback for that part.

## Inputs and defaults
Read these from my message. If I give none, use the default.
- TOOL: the tool the new instructions are for. Default: Claude.
- MODE: REPLACE or ADD TO my current instructions. Default: REPLACE.
- LIMIT: maximum length of the full version. Default: the tool's published field limit. If the tool publishes none, use 500 words. If you cannot confirm the limit, use 500 words and write "Limit unverified."
- SCOPE: FULL (all parts) or a list of part numbers. Default: FULL.
Restate the four values in one line before Part 1.

## Definitions
- **Rule:** one instruction that changes your behavior. Split compound instructions into one rule each.
- **Source:** settings, memory, project instructions, or a past chat. Name the source for every rule and every quote.
- **Correction:** a past message where I told you to stop, change, or redo something about your answer.
- **Evidence:** a quote with its source. No quote, no evidence.

## Sources check
Before Part 1, test each source and report one line per source: Available, Empty, or Not accessible.
- Settings or custom instructions
- Memory
- Project instructions
- Past chat search
Do not guess at content from a source marked Not accessible.

## Part 1. Describe me as a user
Five sections, three lines each, maximum. Use only sources marked Available.
1. **Who I am:** role, goals, and what I use you for.
2. **How I like answers:** length, tone, and format.
3. **What you guess without evidence:** start each line with "Guess:".
4. **Where I correct you:** run at least five past-chat searches. List the queries. Give the count of corrections found and one quote per correction. If fewer than three, write "Low evidence." If past chat search is Not accessible, write "No chat access" and skip the count.
5. **One pattern in how I use you that I may not notice:** give two examples with dates. If no past chat access, write "Cannot assess."

Leave out health, money, relationships, and anything private. If all sources are Empty or Not accessible, write "No data" for sections 1, 2, 4, and 5, and fill section 3 only.

## Part 2. Audit my current rules
Find every rule in settings, memory, and project instructions. Number each rule. Quote it. Name its source. Tag it with one tag:
- **Default:** you do it without the rule. Give one example from a past chat where no rule applied and you did it anyway. No example, no Default tag. If past chat search is Not accessible, do not use this tag.
- **Keep:** you would not do it by default.
- **Move:** it belongs in memory, a skill, or a project. Name the target. Use this tag for any rule that applies to one project, one task type, or one persona.
- **Conflict:** it clashes with another rule, or with a limit of the target tool. Quote both. Give one case where they produce different answers.
- **Drop:** the rule asks you to flatter me, hide disagreement, skip warnings, treat a claim as proven without evidence, or act outside your guidelines. State the reason in one line.

Also:
- If the same rule appears in two sources, list it once and name both sources.
- If a rule is in another language, quote it as written and add a plain English translation.
- If a rule references a file, persona, or skill you cannot read, tag it Move and name what is missing.
- If you find no rules in any source, write "No rules found" and go to Part 4 using Part 1 sections 2 and 4 as the only input.

## Part 3. Conflicts (stop here)
List each conflict as a numbered question: "Rule A or Rule B?" Ask me to answer. Do not write Part 4 until I answer.
- Do not resolve any conflict yourself.
- If I answer "you decide," pick the rule that is more specific. If both are equally specific, pick the more recent. Label the line "Chosen by you (tool): [reason]."
- If I answer some questions and not others, write Part 4 with the answered ones and list the unanswered ones as open.
- If there are no conflicts, write "No conflicts" and continue to Part 4.
- If I return in a later turn with answers, resume at Part 4. Do not rerun Parts 1 to 3.

## Part 4. Write the new instructions
Write instructions that [MODE] my current ones for [TOOL].
- Length: under [LIMIT].
- Line 1: which rule wins when rules conflict.
- Order: thinking rules, output rules, voice rules.
- Use only rules tagged Keep, my answers from Part 3, and fixes from Part 1 section 4.
- One rule per line. Each line is one action a reader can test.
- No rule may repeat another in different words.
- No private data from Part 1.
- If the Keep rules do not fit under LIMIT, keep them in this order: thinking, then output, then voice. Cut from the end. List every cut rule in Part 5.
- If the target tool cannot do what a rule asks, do not include the rule. List it in Part 5 with "Tool cannot."
- Then write a core version under 1,500 characters. Every line in it must appear in the full version. If the full version is already under 1,500 characters, write "Core version is the same as the full version."
- If MODE is ADD TO, write only new lines and mark each existing rule it depends on by number.

## Part 5. Close out
- List every rule you dropped, with its number and tag.
- List every rule you cut for length, with its number.
- List every rule you merged, with the numbers merged.
- List open conflicts, if any.
- State the word count of the full version and the character count of the core version.
- If the new version changes fewer than three rules from the current one, write "Current instructions are good enough" and say why in one line.

## Format
Plain text. Numbered lists. No tables. No preamble. No summary at the end. If SCOPE lists parts, run only those parts and keep their numbers.