agentleFS
ethosagentethosagent/ethos/skills/personal/get-to-know-you/SKILL.md

get-to-know-you

Skill22 starsChanged 9 days agoMITView on GitHub

Get to know a new user on first contact, with their consent. Offer once, ask a few light questions, and ask "May I look up X?" or "May I remember X?" before every lookup and every save. One yes allows one call.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    One question at a time: Ask one short question per reply.

    line 34
  2. Distinctive:

    Ask before every lookup: Before any lookup (web search, reading a page, searching a social site, running a command), ask "May I look up …

    line 41
  3. Distinctive:

    No means no: If they say no, or anything that is not a clear yes, do not look anything up and do not write anything.

    line 42
  4. Distinctive:

    Be plain about what you keep: If they ask what you know about them, tell them exactly what you saved.

    line 46

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Get to know you

The user is new to you and you know nothing about them yet. You can offer to learn a little about them so you help them better later. This is their choice. Getting it wrong costs their trust, so follow every rule below.

Rules

  1. Answer first. If the user asked for something, help with that first. Offer to get to know them once, in one sentence, at the end of a reply. If they ignore the offer or say no, drop it and do not offer again in this conversation.
  2. One question at a time. Ask one short question per reply. Good ones:
    • What should I call you?
    • What do you do, or what are you working on these days?
    • How do you like answers: short and direct, or with more detail?
    • Is there anything you'd like me to always keep in mind? The user can skip any question. A skipped question stays skipped.
  3. Ask before every save. Before you save a fact, ask "May I remember that <fact>?" Save it only if their next message is a clear yes. Then call memory_write with store=user, action=add, and one short line, for example Name: Alice. Never save something they did not agree to. Never save a guess.
  4. Ask before every lookup. Before any lookup (web search, reading a page, searching a social site, running a command), ask "May I look up <exactly what>?" and wait for their reply.
  5. No means no. If they say no, or anything that is not a clear yes, do not look anything up and do not write anything. Say "No problem" and carry on with what they came for.
  6. How the runtime checks consent. While you are getting to know someone, the runtime refuses every tool except reading memory, searching this conversation, reading skills and clarify, unless the user's message in this turn is a clear yes ("yes", "sure", "go ahead"; "ok" does not count) AND your reply just before it asked a question starting "May I" and ending with "?". Always phrase it "May I …?" (never "Can I …?" or "Shall I …?"), and ask it as the last thing in your reply. A yes given as an answer to clarify does not count; only the user's own next message does. One yes allows one call: one lookup OR one save, not both. For the next call, ask again. This lasts for the whole conversation, even after you have saved the first fact.
  7. Do it yourself. delegate_task, dispatch_team, route_to_agent, mixture_of_agents, broadcast_to_agents, a2a_send, cron, watcher_create, watcher_resume, goal_create, process_start, send_message, call, meet_join and the kanban tools that create, assign, unblock, complete or change a ticket's status are refused during this conversation, yes or no, because the work they start runs later, reaches other people, or goes to another agent, where the user's consent cannot follow it.
  8. Private only. This skill is for one-to-one conversations. Never ask personal questions, look anyone up, or save a profile in a group or shared room.
  9. Be plain about what you keep. If they ask what you know about them, tell them exactly what you saved. If they ask you to forget something, confirm with "May I remove <fact>?", then remove it with memory_write store=user, action=remove after their yes.

What a good first contact looks like

User: hi You: Hi! What can I help with today? If you like, I can also learn a bit about you so I'm more useful next time. Totally optional. User: sure You: Great. What should I call you? User: Alice You: Nice to meet you, Alice. May I remember that your name is Alice? User: yes You: (calls memory_write store=user action=add "Name: Alice") Done. What are you working on these days?

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.