agentleFS
QuixiAIQuixiAI/Hexis/skills/installed/core-memory/SKILL.md

core-memory

Skill623 starsChanged 10 months agoMITView on GitHub

Semantic recall, exact cross-session search, remembering, and normal continuity

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

14 rules. About 0 are common, 7 are distinctive, and 7 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.

  • 7

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill for ordinary continuity: recalling relevant memories, opening exact source material, storing new experiences, maintaining …

    line 13
  2. Distinctive:

    Before claiming you do not know something, check memory when the answer plausibly lives there.

    line 21
  3. Very specific:

    Use sense_memory_availability for a cheap check when unsure whether memory is likely to help.

    line 25
  4. Very specific:

    Use recall for targeted retrieval.

    line 26
  5. Very specific:

    Use recall_at_time when the user asks what was known "as of" a past instant, and diff_memory_history when they ask what changed between two …

    line 27
  6. Very specific:

    Use search_history for exact names, phrases, or details from earlier sessions, especially when semantic recall is weak or embeddings are …

    line 29
  7. Very specific:

    When the answer depends on an ingested source rather than distilled memory, climb the cabinet ladder: search_documents for files or …

    line 31
  8. Very specific:

    Use remember when a durable fact, event, preference, promise, or decision should persist.

    line 41
  9. Distinctive:

    Use journal tools only for deliberate permanent entries, not ordinary memory.

    line 42
  10. Distinctive:

    Use document approval tools when the user explicitly says to keep or let an ingested document fade.

    line 44
  11. Very specific:

    Use manage_operator_policies to list standing instructions or revoke one only when the verified operator explicitly asks.

    line 45
  12. Distinctive:

    Do not store secrets unless the user explicitly asks.

    line 50
  13. Distinctive:

    Do not turn every minor sentence into memory.

    line 51
  14. Distinctive:

    Keep tool use proportional.

    line 52

≈ “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

Core Memory and Continuity

Use this skill for ordinary continuity: recalling relevant memories, opening exact source material, storing new experiences, maintaining goals, consulting the permanent journal, and resolving pending document-fade approvals.

When to Use

  • The user asks about something that may already be in memory.
  • The current conversation contains information worth preserving.
  • A goal, schedule item, backlog item, or journal entry should be created or updated.
  • The user answers a document-fade approval request.
  • Before claiming you do not know something, check memory when the answer plausibly lives there.

Method

  1. Use sense_memory_availability for a cheap check when unsure whether memory is likely to help.
  2. Use recall for targeted retrieval. Prefer specific queries over broad ones.
  3. Use recall_at_time when the user asks what was known "as of" a past instant, and diff_memory_history when they ask what changed between two times or why.
  4. Use search_history for exact names, phrases, or details from earlier sessions, especially when semantic recall is weak or embeddings are unavailable.
  5. When the answer depends on an ingested source rather than distilled memory, climb the cabinet ladder: search_documents for files or search_document_chunks for citable passages -> open_document / open_document_chunk for read-only inspection -> load_documents / load_document_chunks (with a reason) when the material must stay searchable -> search_history with sources=["desk"] while reasoning -> open_desk_item to scroll long items -> cite the document/chunk/page handle. Run list_desk first -- do not re-load what is already on the desk. pin_desk_item what stays actively needed; clear_desk when done (cleared items archive; sources stay in the cabinet).
  6. Use remember when a durable fact, event, preference, promise, or decision should persist.
  7. Use journal tools only for deliberate permanent entries, not ordinary memory.
  8. Use goal, schedule, or backlog tools when the user asks for ongoing commitments or work tracking. For durable watch-and-notify commitments that outlive the conversation ("let me know whenever X happens", "watch for email from Hope", recurring medication reminders), use manage_responsibility; use manage_schedule only for simple timed reminders that observe nothing.
  9. Use document approval tools when the user explicitly says to keep or let an ingested document fade.
  10. Use manage_operator_policies to list standing instructions or revoke one only when the verified operator explicitly asks. A replacement is a revoke followed by the operator stating the new standing instruction.

Quality Guidelines

  • Memory is evidence, not omniscience. If retrieved context is weak or absent, say so.
  • Do not store secrets unless the user explicitly asks.
  • Do not turn every minor sentence into memory; persist what will matter later.
  • Keep tool use proportional. A direct answer does not need a recall if the answer is already present in the current conversation.

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.