Guide · Agent skill
How to use the remember agent skill: save knowledge to memory
Updated · written with AI from the project's own documentation and checked against our catalogue data
The remember skill lets you explicitly save an important fact, convention or lesson to your agent's auto-memory so it is loaded at the start of later sessions in the same project. It is aimed at developers who keep re-explaining the same build command, quirk or debugging gotcha. After this guide you will know what the skill does, how to install it from the ZIP, how to trigger it with plain requests, and what it needs and does not do.
Remember Knowledge Explicitlyalirezarezvani/claude-skills · MIT Download the skill Quick start
- 1Download the ZIP. Get the remember skill ZIP from this page. It contains SKILL.md.
- 2Unzip into a skills folder. Unzip into ~/.claude/skills/ for personal use, or into .claude/skills/ inside a project to scope it to that repository.
- 3Or upload in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the ZIP.
- 4Ask the agent to remember something. Say 'remember this' followed by a concrete fact, such as a build command or a version constraint.
- 5Review the duplicate check. If a similar entry exists, the agent shows it and asks whether to update it or add a new one. Choose the option that fits.
- 6Confirm the saved entry. Check the confirmation showing the entry and the MEMORY.md line count. Future sessions in this project will load it.
What the remember skill makes the agent do
Agent skills are instruction packages that extend what an agent such as Claude Code can do. This one is called remember, and its purpose is to write a deliberate entry into auto-memory when something is important enough that you do not want to rely on the agent noticing it on its own.
When you ask it to remember something, the agent first extracts the concrete fact, why it matters if you said so, and whether it is project-specific or global. It then checks MEMORY.md for near-duplicate entries, appends a short note, and confirms the save. Entries are meant to be one line, and they do not need timestamps or IDs.
- Writes a concise note into the project's MEMORY.md file under your Claude config folder.
- Checks for similar existing entries and asks whether to update or add a new one.
- Warns you when MEMORY.md approaches its line limit.
- Suggests moving strict rules into CLAUDE.md when the note reads like a rule.
Install the remember skill from the ZIP
Download the ZIP for the remember skill from this page. The skill is a single file, SKILL.md, inside the engineering-team/self-improving-agent/skills/remember folder of the alirezarezvani/claude-skills repository, so you can use the download provided here.
There are two ways to install it. For a personal setup, unzip it into your user skills folder at ~/.claude/skills/. To limit it to one repository, unzip it into the .claude/skills/ folder inside that project instead. Alternatively, in the Claude apps you can upload the ZIP under Settings → Capabilities → Skills.
- Personal install: unzip into ~/.claude/skills/ so it is available everywhere.
- Project install: unzip into .claude/skills/ in the repository.
- Claude apps: upload the ZIP under Settings → Capabilities → Skills.
Trigger the skill with natural requests
You do not need to type the command. The skill is meant to activate when you say things like 'remember this', or when you ask the agent to capture a lesson so it stops repeating a mistake across sessions. It also applies when a discovery is too important to rely on auto-capture.
If you prefer explicit invocation, the source documents the form /si:remember followed by what to remember. Quote the note so the whole sentence is captured as one entry.
- 'Remember this: the build needs Node 20 LTS, v22 breaks it.'
- 'Capture a lesson: the CDN causes the CORS errors on /api/upload, not the backend.'
- 'Remember that we use barrel exports in src/components/.'
/si:remember "This project's CI requires Node 20 LTS — v22 breaks the build"Examples of good entries
The best entries are concrete. The source's guidance is to include the actual command or value rather than just the concept, and to keep each note to one line. Entries that describe a tool-specific gotcha or an architecture decision tend to be the most useful to a future session.
Preferences you want the agent to learn also qualify, such as a preferred error-handling style or a rule against adding comments that explain obvious code.
- Good: "Build with `pnpm build`, tests with `pnpm test:e2e`".
- Good: "Jest needs `--forceExit` flag or it hangs on DB tests".
- Good: "We chose Drizzle over Prisma for type-safe SQL".
- Weak: "The project uses pnpm for building and testing".
What it needs and how it works under the hood
The skill expects Claude Code or a compatible agent with auto-memory support, and shell access so the agent can run grep against the memory directory. Its duplicate check looks in the project memory folder under ~/.claude/projects/ for the MEMORY.md file. The skill runs scripts, but it does not use the network.
Memory files are plain text, so anything you save is readable by anyone with access to that folder.
- Requires Claude Code or a compatible agent with auto-memory support.
- Requires shell access so grep can search the memory directory.
- Runs scripts: yes. Uses network: no.
- Ships as a single file: SKILL.md.
MEMORY_DIR="$HOME/.claude/projects/$(pwd | sed 's|/|%2F|g; s|%2F|/|; s|^/||')/memory"
grep -ni "<keywords>" "$MEMORY_DIR/MEMORY.md" 2>/dev/nullSafety notes: secrets, scripts and what not to save
Do not use this skill to store credentials, tokens, keys or other secrets. The skill's own guidance says never to save them, and a plain-text memory file is not a safe vault. Use your secret manager or environment configuration instead.
Use other mechanisms for other kinds of knowledge. Temporary context belongs in the conversation itself or session memory. Strict rules that must be enforced are better written to CLAUDE.md through /si:promote, and cross-project rules belong in ~/.claude/CLAUDE.md.
- Never store credentials, tokens or secrets in memory.
- Temporary context: keep it in the conversation.
- Enforced rules: use /si:promote to write to CLAUDE.md.
- Global rules across projects: use ~/.claude/CLAUDE.md.
What people use it for
- Recording the exact build and test commands for a repository so the agent stops asking for them.
- Keeping a version constraint, such as a required Node LTS release, that caused a broken build before.
- Saving a debugging gotcha, for example a test runner flag needed to stop a hang.
- Noting an architecture decision, such as a chosen ORM, so later changes respect it.
- Capturing a coding preference the agent should follow in this project.
Frequently asked questions
Is the remember skill free to use?
The skill is published under the MIT licence in the alirezarezvani/claude-skills repository, as stated on this page. You still need Claude Code or a compatible agent with auto-memory support to run it.
What do I need to run the remember skill?
You need Claude Code or a compatible agent with auto-memory support, plus shell access so the agent can run grep against the memory directory. The skill runs scripts but does not use the network.
Where does the remember skill save my notes?
It appends a concise line to MEMORY.md in the project's memory folder under ~/.claude/projects/. That file is loaded at the start of later sessions in the same project.
How do I install the remember skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use or into .claude/skills/ within a project. In the Claude apps, upload the ZIP under Settings → Capabilities → Skills.
Can I use remember to store API keys or passwords?
No. The skill says never to save credentials, tokens or secrets, and a plain-text memory file is not a safe vault. Keep secrets in a dedicated secret manager or environment configuration.
What is the difference between remember and CLAUDE.md?
Remember writes short notes to auto-memory. CLAUDE.md holds enforced rules with higher priority. The skill suggests promoting rule-like notes to CLAUDE.md, and you can use /si:promote to write directly to it.
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 keep re-explaining the same project quirk, build command or debugging lesson to your agent and want it kept in one repository. Skip it for secrets, for temporary context that belongs in the conversation, or if you do not run Claude Code or a compatible agent with auto-memory support.
Download the skill