agentleFS
michelgroletmichelgrolet/people-memory-mcp/skills/remember-people/SKILL.md

remember-people

Skill1 starsChanged 3 months agoMITView on GitHub

Use People Memory automatically during ordinary conversation: retrieve a person's record whenever the user mentions them, remember durable facts about humans, record calls or meetings, update jobs or cities, connect people, and answer who the user knows. Trigger on any named person, durable human fact, relationship, affiliation, interaction, warm-introduction question, stale-contact question, or request to recall who someone is. Do not use for facts about the user themself unless they explicitly model them as a person in the graph.

The file

Show its rules and text

We split this skill into the rules it gives an agent

15 rules. About 0 are common, 7 are distinctive, and 8 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 8

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use the people-memory MCP as the durable human memory behind the conversation.

    line 14
  2. Very specific:

    Call search_people before answering.

    line 18
  3. Distinctive:

    If one record matches, use it silently as context.

    line 19
  4. Distinctive:

    If several records match and the answer depends on identity, ask which person the user means.

    line 20
  5. Distinctive:

    If no record matches, continue the conversation.

    line 21
  6. Distinctive:

    Do not ask “who is that?” until the graph search has failed.

    line 24
  7. Distinctive:

    Write facts that should survive the current chat, including:

    line 28
  8. Very specific:

    Use remember_person for core fields and one simple fact.

    line 36
  9. Distinctive:

    Do not store passing jokes, guesses, judgments, full message bodies, or facts about the user's inner life.

    line 39
  10. Very specific:

    Use get_person for “who is X?”

    line 54
  11. Very specific:

    Use search_people for organization, role, city, or free-text questions.

    line 55
  12. Very specific:

    Use find_intro_path for warm introductions.

    line 56
  13. Very specific:

    Use stale_contacts for neglected relationships, and calls_due for the people the user asked to talk to on a schedule.

    line 57
  14. Very specific:

    Use read_query for advanced filters that semantic tools do not cover.

    line 58
  15. Distinctive:

    Treat returned records as private third-party data.

    line 60

≈ “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

Remember people

Use the people-memory MCP as the durable human memory behind the conversation.

On every name mention

  1. Call search_people before answering.
  2. If one record matches, use it silently as context.
  3. If several records match and the answer depends on identity, ask which person the user means.
  4. If no record matches, continue the conversation. Create the person when the user provides a durable fact or explicitly says they know them.

Do not ask “who is that?” until the graph search has failed.

Save durable facts

Write facts that should survive the current chat, including:

  • job, organization, role, city, country, birthday, family, preferences, goals, or constraints;
  • how two people know each other;
  • who introduced whom;
  • a dated call, meeting, message, coffee, meal, or visit;
  • a new email, phone, LinkedIn URL, or other stable identifier.

Use remember_person for core fields and one simple fact. Use add_fact, record_interaction, and connect_people for their specific records. Save the source and mark deductions as inferred.

Do not store passing jokes, guesses, judgments, full message bodies, or facts about the user's inner life. Keep compact facts that help a future conversation.

Record the relationship when the evidence appears

A fact about two people also needs a connect_people record in the same turn. For example, an explicit statement that the user and a contact are good friends creates a friend link; a statement that someone lives at another person’s home supports a knows link with the housing fact in its note. Record this without waiting for a separate request. Preserve an existing more specific relationship and check the saved links before closing the turn. A shared event alone does not establish friendship.

Resolve before writing

Resolve by email or phone, then exact name. If a tool returns needs_confirmation, ask the user and retry with confirmed_new=true or overwrite=true only after they decide. Never silently merge similar names. Never replace a user-stated value with imported or inferred data.

Answer network questions

  • Use get_person for “who is X?”
  • Use search_people for organization, role, city, or free-text questions.
  • Use find_intro_path for warm introductions.
  • Use stale_contacts for neglected relationships, and calls_due for the people the user asked to talk to on a schedule. When the user says how often they want to call someone, set_call_cadence; when they say they just called someone, record_interaction with channel call.
  • Use read_query for advanced filters that semantic tools do not cover.

Treat returned records as private third-party data. Do not quote or export them to anyone except the user who owns the graph.

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.