How do I install the Memory Guide skill?

Run npx skillmds@latest add otodock/memory-guide in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Memory Guide safe to use?

This skill has not completed SkillMD's automated safety review yet. Capability flags: makes network calls. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Memory Guide?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Memory Guide free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Memory Guide?

OtoDock (@otodock) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Memory Guide by otodock — Agent Skill

Name: Memory Guide
Rating: 4.666666666666667 (3 reviews)
Author: OtoDock

File contents

Memory tool — worked examples

The memory-usage card in your instructions carries the rules (scopes, what not to save, revise instead of duplicate). These are the command shapes behind them. Memory is markdown topic files: /memories/agent/ (shared with every user of this agent) and /memories/user/ (private to the current user). A generated MEMORY.md index per scope is maintained by the platform — never edit it; edit topic files and the index follows.

Save a new fact (create a topic)

memory(
  command="create",
  path="/memories/user/preferences.md",
  file_text="# Communication preferences\n- Prefers replies in Greek (2026-06-12)\n- Wants metric units everywhere (2026-06-12)\n"
)

Start every topic file with a one-line # heading — it becomes the topic's index entry. Date each fact (YYYY-MM-DD) so staleness stays visible. create errors if the file exists — that's your cue the topic already exists: UPDATE it instead of duplicating.

Update an existing fact (revise in place)

memory(
  command="str_replace",
  path="/memories/user/preferences.md",
  old_str="- Wants metric units everywhere (2026-06-12)",
  new_str="- Switched to imperial units (2026-07-02; was metric until then)"
)

old_str must match exactly once. Supersede outdated facts with a short "was X until DATE" trail instead of silently erasing them — it keeps the history readable for you and the user. Prefer targeted str_replace / insert edits; never rebuild a whole topic from scratch when a small edit will do.

Append to a topic

memory(
  command="insert",
  path="/memories/agent/post-history.md",
  insert_line=2,
  insert_text="- Posted launch teaser to company Instagram (2026-06-12)"
)

Text is inserted AFTER the given line (insert_line=0 = top of file). Use view first if unsure of the layout.

Remove a wrong or dead memory

memory(command="delete", path="/memories/user/old-project.md")

Delete topics that are wrong or no longer matter — stale memories are worse than no memories. Everything is git-versioned platform-side, so deletion is recoverable by humans.

Read a topic that wasn't inlined

When a scope outgrows the inline budget, your prompt carries only its index. Fetch a topic on demand:

memory(command="view", path="/memories/agent/infrastructure.md")

view on a directory lists it: memory(command="view", path="/memories/agent").

---name: memory-guidedescription: Worked examples for the memory tool — create a topic, revise a fact in place, append, delete, view a topic that was not inlined, and the exact command shapes. Load when unsure how to write or edit a memory topic file.---# Memory tool — worked examplesThe `memory-usage` card in your instructions carries the rules (scopes,what not to save, revise instead of duplicate). These are the commandshapes behind them. Memory is markdown topic files: `/memories/agent/`(shared with every user of this agent) and `/memories/user/` (private tothe current user). A generated `MEMORY.md` index per scope is maintained bythe platform — never edit it; edit topic files and the index follows.## Save a new fact (create a topic)```memory( command="create", path="/memories/user/preferences.md", file_text="# Communication preferences\n- Prefers replies in Greek (2026-06-12)\n- Wants metric units everywhere (2026-06-12)\n")```Start every topic file with a one-line `# heading` — it becomes the topic'sindex entry. Date each fact `(YYYY-MM-DD)` so staleness stays visible.`create` errors if the file exists — that's your cue the topic alreadyexists: UPDATE it instead of duplicating.## Update an existing fact (revise in place)```memory( command="str_replace", path="/memories/user/preferences.md", old_str="- Wants metric units everywhere (2026-06-12)", new_str="- Switched to imperial units (2026-07-02; was metric until then)")````old_str` must match exactly once. Supersede outdated facts with a short"was X until DATE" trail instead of silently erasing them — it keeps thehistory readable for you and the user. Prefer targeted `str_replace` /`insert` edits; never rebuild a whole topic from scratch when a small editwill do.## Append to a topic```memory( command="insert", path="/memories/agent/post-history.md", insert_line=2, insert_text="- Posted launch teaser to company Instagram (2026-06-12)")```Text is inserted AFTER the given line (`insert_line=0` = top of file). Use`view` first if unsure of the layout.## Remove a wrong or dead memory```memory(command="delete", path="/memories/user/old-project.md")```Delete topics that are wrong or no longer matter — stale memories are worsethan no memories. Everything is git-versioned platform-side, so deletion isrecoverable by humans.## Read a topic that wasn't inlinedWhen a scope outgrows the inline budget, your prompt carries only its index.Fetch a topic on demand:```memory(command="view", path="/memories/agent/infrastructure.md")````view` on a directory lists it: `memory(command="view", path="/memories/agent")`.

Frequently asked questions

Memory Guide