skills-for-selfdev

Skills for Selfdev

Using Claude Code? Build your daily assistant: copyable prompt and component map.

AI skills for self-development. Turn experience into understanding, and understanding into a next step you choose.

Русский · Install · All skills · Privacy

You finish a conversation, close a lecture, or reach the end of a busy day. There is something worth keeping—but it is scattered across a transcript, a few notes, and your memory.

This collection gives your AI assistant a repeatable way to work with that material: preserve what happened, explore what it might mean, and help you choose what to try. Each skill includes instructions; some also include templates, scripts, or a local runtime.

Start small: bring one transcript or a short account of your day. You do not need to build a personal knowledge system first.

Choose an installation route

The unified INSTALL guide covers requirements, commands, success checks, updates, and disabling each route. The routes are independent: Telegram and computer history are optional. A Cursor collector and automatic import of its runtime into Telegram are not bundled.

Cursor and Codex in evening and weekly reviews: keep human and machine time separate and preserve coverage gaps.

One practice or a personal system

For example: reflect on a consultation → keep an accepted understanding → choose an action → review the outcome at the end of the week. Evening Ten offers further possibilities; the daily assistant supports the journal and schedule. Each module also works on its own.

Find your starting point

Try one skill

You need Git and Python 3.10+. Clone the repository, then copy only the skill you want into your agent's skill directory:

git clone https://github.com/KiteTools/skills-for-selfdev.git
cd skills-for-selfdev
python3 scripts/install.py learning-notes

The installer copies files to ~/.agents/skills/. It does not start a bot, connect an account, schedule messages, or upload your documents. It refuses to replace an existing installation.

In Codex, try:

Use $learning-notes with this transcript. Keep the speaker's examples,
mark unclear passages, and write the result in my language.

For Claude Code, select your project's skill directory with --dest: command and first trial. For another agent, use its supported skill location or give it the selected SKILL.md and its referenced files. Tool-dependent steps still require the tools named by that skill. Installation routes and requirements →

From an episode to a next step

Notice. Begin with an actual episode, source, or observation. Understand. Keep facts, reported experiences, and interpretations separate. Try. Choose an action that fits what you understood. Review. Look for an observable result, including a result that disproves the original idea.

The workflows grew out of personal learning and IMT practice. Method-specific ideas are named where they matter; they are not presented as universal scientific findings. The skills support reflection and preparation for a human conversation. They do not establish diagnoses, medical causality, or a person's hidden motives.

What is included—and what connects separately

• Portable workflows: reflection and learning skills work from supplied material with an AI assistant and the tools each task needs. The expert-analysis workflow includes Python helpers; SVG and image outputs need a rendering tool.
• Activity review: an offline CLI processes normalized JSON, with separate human and machine time. Export, normalization, and collectors are configured separately.
• A local runtime: Personal Daily Assistant includes an installable OpenClaw/Telegram package. Its bundled snapshot is documented; account setup and live verification happen on your own machine.
• An external app workflow: Longevity Intake connects to Longevity; the application is maintained in its own repository.
• Separate applications: Building Your Life Function provides a local life-event timeline; Personal Cabinet provides application source and setup for your own accounts. Their integration guides explain the dependencies; hosted services are not bundled here.

Prism is an external writing tool. It is not packaged or reimplemented in this collection.

For the Personal Cabinet's everyday workflow, read the complete user guide; to deploy it, use the separate operator installation guide and SendPulse setup kit. The portability plan distinguishes the implemented source packages from future delivery queues, adapters, and migration work. Live setup on your own accounts still needs verification. All visitor guides and skill descriptions are also available in Russian.

Private by design of the package

This repository contains workflow instructions, original illustrations, empty templates, and clearly fictional examples. It contains no personal diaries, client sessions, family histories, account tokens, or private activity exports.

Your chosen AI provider and connected services still have their own data policies. Installing a skill does not make a cloud assistant local. Read the data boundaries →

About this release

Collection 0.3.3; bundled Telegram assistant runtime 0.3.0. Release 0.3.3 adds review rules for Codex/Cursor machine time; it does not automatically update a running assistant. Requirements and verification boundaries · Validation history · GitHub Actions · Contributing · MIT license

Built by KiteTools. Explore the wider collection of IMT tools and artifacts.

Update each installation route · Assistant foundation changes in 0.3.0 · 0.2.0 history