Guide · Agent skill
How to use the Remember agent skill: organise session knowledge
Updated · written with AI from the project's own documentation and checked against our catalogue data
The remember skill is an agent skill that reviews what your AI agent learned during a working session and decides where each piece of knowledge should be kept. Instead of leaving useful findings buried in chat history, it sorts them into project memory, a notepad, or durable instruction files. This guide explains what the skill does, how to install it, how to ask for it, and where to be careful.
Quick start
- 1Download the ZIP. Get the remember skill ZIP from the skill's page on this site.
- 2Choose an install location. Unzip into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder for one project.
- 3Or upload in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the ZIP.
- 4Finish a working session. Once useful findings have emerged, ask the agent in plain language to save or organise them.
- 5Review the proposed destinations. Check which items go to project memory, the notepad, or AGENTS.md or CLAUDE.md before the agent writes.
- 6Check the summary. Read the list of what was stored, where it went, and any duplicates or conflicts to clean up.
What is an agent skill?
An agent skill is a packaged set of instructions that an AI agent can load when a task matches it. The skill is defined in a file named SKILL.md, which describes when to use the skill and how the agent should work through it.
The remember skill comes from the oh-my-claudecode project (Yeachan-Heo/oh-my-claudecode, MIT licence) and is listed here as skills/remember. Its only file is SKILL.md.
What the remember skill makes the agent do
When you ask the agent to preserve what was learned in a session, the skill tells it to gather the relevant findings and classify each one as a durable project fact, a temporary working note, an operator preference or instruction, or duplicate, stale or conflicting information.
The agent then proposes a destination for each item. The skill names four memory surfaces: project memory for durable team knowledge, a notepad priority store for short high-signal context, a notepad working store for temporary active-session notes, and docs or AGENTS/CLAUDE files for durable instructions and conventions when they truly belong there. It writes only to the appropriate surface and finishes by listing what was stored, where, and any duplicates or conflicts to clean up.
- Do not dump everything into one store.
- Prefer project memory for durable team knowledge.
- Prefer the notepad for short-lived working context.
- Keep entries concise and actionable.
- If something is uncertain, mark it as uncertain rather than storing it as fact.
How to install the remember skill
Option 1: download the ZIP from the skill's page on this site and unzip it into your personal skills folder, ~/.claude/skills/. The skill folder should contain SKILL.md.
Option 2: unzip it into a project's .claude/skills/ folder so that only that project uses the skill.
Option 3: in the Claude apps, upload the ZIP under Settings → Capabilities → Skills.
How to trigger the skill with natural requests
You do not need a special command. The skill is meant for requests where you want session knowledge preserved or organised. Describe what you want kept in plain language, and mention the outcome you expect.
Good requests name the kind of result: that you want findings saved, sorted, or checked for conflicts. The skill also covers the case where you want existing notes reviewed for stale or contradictory entries.
- Save what we learned about the test setup in this session to the right memory place.
- Organise the findings from today's debugging into project memory and notepad as appropriate.
- Review our existing convention notes and flag anything stale or conflicting.
- Preserve the decisions we made about the API design so the next session has them.
Examples of what the agent will do
Suppose a session uncovered that the build needs a specific environment variable, the team prefers a particular naming pattern, and a debugging workaround only matters for the next few turns. Under this skill, the agent should classify these separately: the build requirement and naming convention are durable project facts, and the workaround is temporary working context.
The skill also asks the agent to mark uncertain items rather than store them as fact. If a finding conflicts with a note already in memory, the agent should call out the conflict in its summary so you can resolve it.
What the skill needs and whether it runs code
The item page lists no requirements. The skill runs no scripts and makes no network calls, according to the listing. Its single file, SKILL.md, is a set of written instructions for the agent.
Safety notes before you use it
Review what is written. The agent may store uncertain items or put them in the wrong place, and unchecked writes can clutter or mislead future sessions. Read the proposed destinations and the summary of what was stored.
Keep secrets out of memory. The skill does not explain how sensitive content is filtered, so do not put passwords, keys or other sensitive data into memory files.
Skip it for one-off tasks. If nothing needs to persist, the memory writes add clutter with no lasting benefit.
Tips for getting good results
Ask for the skill at the end of a session, when the useful findings are clear. Keep entries short, and ask the agent to mark anything it is unsure about. For team projects, review convention files before they are committed so that they stay concise and accurate.
- Ask for conflicts to be listed, not silently resolved.
- Periodically ask for a review of stale notes.
- Keep the memory files under version control if your team shares them.
What people use it for
- Keeping build and test requirements in project memory so the next session starts with them.
- Storing team naming conventions or architecture decisions in a durable instructions file.
- Holding short-term debugging context in the notepad for the next few turns.
- Reviewing existing memory notes for stale or contradictory entries before they pile up.
Frequently asked questions
How do I install the remember skill?
Download the ZIP from this site's skill page and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder. In the Claude apps you can instead upload the ZIP under Settings → Capabilities → Skills.
Does the remember skill need anything else installed?
The listing states no requirements. It runs no scripts and uses no network, and its only file is SKILL.md.
Is the remember skill free?
The source is published under the MIT licence, which is stated on the item page. This page does not list prices, so check the source repository for licence terms.
What is the difference between project memory and the notepad?
Project memory is for durable team or project knowledge. The notepad holds short-lived working context, with a priority store for high-signal items for the next turns and a working store for temporary session notes.
Can the remember skill store secrets?
It is not recommended. The skill does not say how sensitive content is filtered, so keep passwords, keys and other sensitive data out of memory files.
What should I do if the agent stores something wrong?
Review the summary of what was stored and where, then correct or remove the entry in the relevant file. The skill itself asks the agent to mark uncertain items rather than store them as fact.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use the remember skill if you work with an AI agent over many sessions and want project knowledge kept outside chat history, or if your team maintains shared convention files that need to stay accurate. Skip it for one-off tasks, or if you cannot review what gets written to memory, and keep secrets out of it.
Download the skill