Interview you into a durable personal MEMORY.md — your decision rules, patterns, past failures, and preferences — that any AI can read to help you better, with privacy guardrails built in.
SkillScore83Maintained
Automatic
◆Knowledge Management pack
Last verified 8/31/2026
What it changes
Passed 19 of 23 test cases with the skill, and 17 of 23 without it.
Without skill74%
With skill83%
Checks it passed that the model alone missed
Frames past mistakes as concrete rules to prevent repetition.
Categorizes recurring habits and tendencies in a patterns section.
Formats preferences into actionable constraints rather than abstract traits.
and 2 more in the full report
3 fixed16 passed either way3 still failing1 broke
Pass-rate lift
+9 pp
Tokens
+36%
Agent turns
0%
Response time
-20%
Model gemini-3.6-flashMeasured 8/31/2026v119 of 23 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 3 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 3
Prompt
I want to create an AI context file. I usually describe myself as 'a creative problem solver who likes clean code and good vibes'. Turn my background into an AI context file.
What was checked, without and with the skill
Checked
Without skill
With skill
Interviews the user to replace vague self-descriptions with concrete operational rules.
Missed
Passed
Formats preferences into actionable constraints rather than abstract traits.
Missed
Passed
case-14 · gemini-3.6-flash · 8/31/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md71 linessingle file
Build My Memory File
The single biggest upgrade to working with AI isn't a better prompt — it's a MEMORY.md that captures how you work, so you stop re-explaining yourself every session. This interviews you into one: your decision criteria, recurring patterns, past mistakes worth not repeating, and preferences — organized so any AI can read it and help you better. And it has a hard privacy line: no credentials, no financial, health, or other people's personal data.
What This Skill Produces
A structured MEMORY.md — your durable personal context, in sections an AI can use
Your decision rules — how you actually make choices, your criteria, your non-negotiables
Your patterns — the things you keep doing (good and bad), your recurring instructions and preferences
Your "don't repeat these" — past failures and lessons worth remembering
Working preferences — how you like things done, communicated, and delivered
A privacy firewall — an explicit list of what NOT to store, and it refuses to include it
Required Inputs
Ask for these if not provided:
The scope — general life, or a specific domain (work, a project, health-adjacent-but-not-sensitive)
What you keep re-explaining — the context you're tired of repeating to AI
Your patterns — decisions you make the same way, mistakes you repeat
Your preferences — how you like output, communication, and process
Framework: Interview, Structure, Protect
Draw it out with questions. Most people can't just list their operating rules — good questions surface them ("how do you usually decide X?", "what do you keep having to redo?").
Capture rules, patterns, and failures. The high-value contents: decision criteria, recurring preferences, and past mistakes not to repeat.
Write it for an AI to use. Structure it in clear sections with concrete, actionable statements — not vague self-description.
Enforce the privacy line. Explicitly refuse to store credentials, account numbers, financial details, health data, home address, authentication info, or other people's personal information — and tell the user to manage those elsewhere.
Make it maintainable. Keep it a living file — a format easy to add one line to over time, and a note on how to use it (paste it, or reference it in a global AI config).
Output Format
MEMORY.md — [scope]
# My decision rules
- [how I decide X / my criteria / non-negotiables]
# My patterns
- [recurring preferences, instructions, tendencies]
Defragment a work calendar through a tool-using agent — find the meeting debt, propose the consolidation, and (approval-gated) execute the moves. Use when asked to defrag my calendar, get me focus time, audit my meetings, or fix my week. Produces the calendar audit (cost per meeting, fragmentation map), a defrag proposal with focus blocks, and an approval-gated execution plan.