How to Work With Me README template (that works for humans and AI).
By Bernard Huang · Updated
A strong 'How to Work With Me' README answers five concrete questions in under one page: (1) how you want information structured, (2) how you prefer feedback and pushback, (3) when to interrupt vs proceed asynchronously, (4) how you handle uncertainty and bad news, and (5) your default blind spots. Build yours interactively at agent-tune.com/tools/how-to-work-with-me or copy one of the dual human + AI Markdown templates below.
On this page
Why one personal README now serves two audiences.
Compare what a new coworker needs on week one with what Claude Code or ChatGPT needs in session one:
| Section | What it tells human teammates | What it tells ChatGPT / Claude / AGENTS.md |
|---|---|---|
| Default format | "Send a 3-line Slack TL;DR above any long doc." | "Lead with the direct conclusion in sentence one before supporting detail." |
| Pushback & disagreement | "Challenge my plan early before writing code." | "Critique weak assumptions before implementing; cite concrete tradeoffs." |
| Autonomy vs. check-ins | "Ship reversible fixes; flag schema changes first." | "Execute reversible steps directly; ask before destructive or cross-boundary changes." |
| Bad news & uncertainty | "Surface blockers on day one, even without a fix yet." | "Flag material uncertainty and failed checks immediately instead of guessing." |
Want to generate yours from structured fields in 60 seconds? Open our free client-side How to Work With Me Card Builder → (exports clean Markdown and a privacy-safe share link).
Template 1: Engineering / IC README (GitHub, Notion & AGENTS.md).
Template 2: Product / Manager README (Slack, Notion & ChatGPT/Claude).
Template 3: The portable AI-ready block (CLAUDE.md, AGENTS.md, Soul.md).
Append this compact companion block straight into ~/.claude/CLAUDE.md, ~/.codex/AGENTS.md, ChatGPT Custom Instructions, or Meta Muse Soul.md:
Not sure which defaults match how you naturally operate? Take our 3-minute free DISC workplace test or build rule-by-rule in the Communication Preferences Builder.
Questions people ask.
What should you leave out of a 'How to Work With Me' README?
Leave out excuses for bad behavior ('I'm blunt so don't take it personally'), rigid personality horoscope labels that box teammates in, and after-hours response expectations disguised as personal quirks. Good READMEs lower friction for collaborators rather than shifting emotional labor onto them.
How long should a personal 'How to Work With Me' document be?
Keep the human version to 150–300 words (one screen in Notion or Slack) and the AI agent version to 6–12 bullet lines so neither teammates nor language models skim past what matters.
Can I generate a 'How to Work With Me' card without taking a personality test?
Yes. Both agent-tune.com/tools/how-to-work-with-me and /tools/communication-preferences-builder build plain-English Markdown directly from your answers—zero personality test or email required.
Where should I save the AI version of my 'How to Work With Me' rules?
Save personal communication defaults globally—in ChatGPT Custom Instructions, Claude Profile Preferences, Meta Muse Soul.md, ~/.claude/CLAUDE.md, or ~/.codex/AGENTS.md—so shared team repos only contain project conventions.
What changed.
- Published with dual human + AI README templates for engineers, managers, and cross-agent instruction files.