Guide · Agent skill
How to use the Custom Memory Mode Creator agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The mode-creator skill is an agent skill for claude-mem, a memory tool that normally writes notes about your coding sessions automatically. It walks the agent through designing a custom mode that records the kinds of notes you care about, with its own observation types and concept tags, then validates, installs and activates that mode. This guide explains what the skill does, how to install it, how to ask for it in plain language, and what to check for safety before you use it.
Custom Memory Mode Creatorthedotmack/claude-mem · Apache-2.0 Download the skill Quick start
- 1Download and unzip the skill. Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use or into .claude/skills/ in a project. In the Claude apps, you can upload it under Settings → Capabilities → Skills instead.
- 2Check the requirements. Confirm you have a local claude-mem worker installation, Node.js 20 or newer, and an interactive question tool in your agent runtime.
- 3Ask for a custom mode in plain language. Describe the work you do and the kinds of notes you want to keep. The agent will begin with an interview rather than asking for JSON fields.
- 4Approve the taxonomy. Review the proposed observation types, concept tags, and skip rules. Rename, add or remove items until the plan fits your needs, then approve it.
- 5Decide on Telegram alerts. Choose whether to receive alerts. If you say yes, select exact types or tags and let the helper handle the token securely.
- 6Verify the active mode. After installation the agent restarts the worker and checks that the active mode appears in startup context. Only trust the result if the installer reports success.
What the mode-creator skill makes the agent do
The skill turns a vague wish such as 'remember the things that matter in my law studies' into a working claude-mem mode. The agent interviews you, proposes a small set of observation types (mutually exclusive kinds of notes) and concept tags (reusable themes you may search for later), and waits for your approval before it installs anything.
Examples from the source include a law student recording case rules, exam traps and conflicting doctrines, or an architect recording design decisions, code constraints and client preferences. The skill can also set up optional Telegram messages for chosen types of notes.
- Interviews you about what is worth remembering and what should be skipped
- Proposes observation types and concept tags for your approval
- Installs the approved mode into claude-mem's data folder and activates it
- Optionally configures Telegram alerts for selected types or tags
- Checks that the active mode appears in startup context
What you need before you install
The skill expects a local claude-mem worker installation, Node.js 20 or newer, and an interactive question tool in the agent runtime. It also needs filesystem access. Telegram setup is separate and only applies if you want alerts.
- A local claude-mem worker installation
- Node.js 20 or newer
- An interactive question tool in the agent runtime
- A Telegram account and bot token (only for optional Telegram alerts)
- Network access (only for Telegram setup)
How to install the skill
Download the ZIP file for the skill from this page and unzip it. Then choose one of two locations. For a personal install, place the unzipped folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.
If you use the Claude apps instead, upload the skill under Settings → Capabilities → Skills.
- Personal install: ~/.claude/skills/
- Project install: .claude/skills/ inside your project
- Claude apps: Settings → Capabilities → Skills
How to trigger the skill with natural requests
You do not need to know the word 'mode'. The skill's description says it applies when someone asks to customize what claude-mem remembers, change a mode, track domain-specific notes, add observation types or tags, or send Telegram notifications for particular memories.
Describe the work you do and what you want to keep track of. For example, you might say that you are a law student and want claude-mem to note every case that establishes a rule, or that you are an architect who wants design decisions and client preferences recorded.
- 'I am a law student. Can claude-mem keep notes on case rules and exam traps?'
- 'Change what claude-mem remembers so it tracks design decisions and site discoveries.'
- 'Add a tag for constraints and send me Telegram messages when a constraint is recorded.'
What happens during a session
The agent asks follow-up questions in small batches and waits for your answers. It inspects existing bundled and user modes first, and it may reuse or adjust a close match. It does not edit the plugin cache or bundled modes. Custom files are installed under the modes/ folder inside claude-mem's data directory.
Before anything is installed, the agent validates the mode without changing anything. A dry-run check looks like this, with placeholders for your own files:
- Each observed item gets exactly one observation type
- An item may get several concept tags
- Existing observations keep their original types; the new mode applies to future notes
- The skill makes timestamped backups and preserves unrelated settings
node <skill-directory>/scripts/install-mode.mjs --mode <temporary-mode.json> --mode-id <parent--custom-id> --dry-runHow Telegram alerts work
Telegram alerts are optional. If you decline them, the skill leaves every Telegram setting unchanged. If you accept, you choose exact observation types and/or concept tags. Matching is OR: once alerts are enabled, a note sends an alert when its type or any of its concepts is selected.
Choosing triggers alone does not turn delivery on. Per-observation alerts default to off, and the skill only enables them with your consent. The bot token is handled through a helper that accepts it through hidden terminal input, so it never needs to appear in chat.
- Alerts include the observation type, title, subtitle, project and observation ID
- Alerts do not include the full narrative or facts
- Avoid selecting categories that may expose sensitive material
- If the agent cannot give you an interactive terminal, it shows you the helper command to run yourself
node <skill-directory>/scripts/configure-telegram.mjs --types <comma-separated-approved-types> --concepts <comma-separated-approved-concepts>Safety notes: scripts and network use
This skill runs scripts and uses the network. The scripts merge settings, validate modes and, if you opt in, configure Telegram. The skill preserves existing settings and existing Telegram triggers, and it writes changes atomically with backups.
The skill will not install a per-user mode on a shared server. If the CLAUDE_MEM_RUNTIME setting is server, it stops before making changes. Do not share your Telegram token in chat, command arguments, logs or tool output. Treat it like a password.
- Runs scripts: yes
- Uses network: yes, only for Telegram setup
- Stops if CLAUDE_MEM_RUNTIME is server
- Never pass the Telegram token as a command argument
What people use it for
- A law student records case rules, exam traps and doctrinal conflicts for later revision
- An architect keeps design decisions, code constraints, client preferences and site discoveries in one searchable place
- An ML platform engineer tracks experiment outcomes, data-contract changes, production incidents and reproducibility risks
- A developer creates a domain-specific variant of the standard code mode and sends selected notes to a Telegram chat
Frequently asked questions
Do I need the standard code mode if I install mode-creator?
Not necessarily. The skill says that if the standard code mode is enough, a custom file is unnecessary, and the agent will offer to keep it. A custom mode is most useful when you want notes tailored to a non-coding or specialized field.
Can I use mode-creator on a shared server?
No, not for per-user modes. The skill stops before making changes if CLAUDE_MEM_RUNTIME is set to server, because it cannot safely install a per-user mode into the shared server.
Will my existing observations change when I install a custom mode?
No. Existing observations keep their original types. The new mode applies only to future observation generation.
Are Telegram alerts sent for every note?
No. Alerts are off by default. After you opt in, only notes whose observation type or concept matches a selected trigger are sent. Alerts include the title and subtitle, so avoid categories that may contain sensitive material.
Where do I put the skill so my agent can find it?
Unzip it into ~/.claude/skills/ for a personal install or into .claude/skills/ inside a project. In the Claude apps, upload it under Settings → Capabilities → Skills.
Do I need Telegram to use the skill?
No. Telegram is optional and only needed for alerts. The skill otherwise needs a local claude-mem worker, Node.js 20 or newer, and an interactive question tool.
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 mode-creator if you run claude-mem and want its notes tailored to a non-coding or specialized field, or want selected memories sent to Telegram. Skip it if the standard code mode already meets your needs, if you work on a shared server, or if you cannot keep sensitive material out of a messaging app.
Download the skill