Skip to content

Skills Guide

Skills are one of the main ways Kabot becomes more capable without turning into a rigid bot.

What A Skill Is

A skill is a structured capability package that can teach Kabot how to approach a task or workflow.

Skills can range from: - planning and debugging workflows - messaging and channel tasks - media and integrations - coding and operational tasks

The Right Mental Model

A skill is not the same as: - a plugin - a model - a bot persona - a shell script

A skill is better thought of as a capability contract: - when it should be used - how the agent should behave - what setup is required - what external tools or APIs it expects

Types Of Skills

In practice, skills usually fall into one of three buckets:

1. Instruction-Only Skills

These mostly steer the agent's behavior and tool usage.

Examples: - planning workflows - debugging workflows - code review workflows

2. Integration Skills

These connect Kabot to real external services, often with: - env keys - OAuth - binaries - API docs

3. Workspace Skills

These are custom capabilities created specifically for your own workspace, team, or domain.

What The Skills Menu Does

In kabot config, the Skills section is for: - enabling or disabling skills - filling in required environment values - understanding which skills need external setup - planning dependency installs

What The Skills Menu Does Not Do

It does not blindly run package-manager installs for everything.

That is intentional.

Kabot separates: - skill definition discovery - skill configuration - dependency planning - actual tool/runtime installation

That separation is healthy because it prevents: - accidental installs you did not mean to run - confusing setup states - broken assumptions about external runtimes already being available

Skill Labels You May See

  • needs env
  • needs binary
  • needs oauth
  • needs node package

These are setup hints, not failures.

What Those Labels Mean In Practice

Label Meaning
needs env the skill expects env values or API keys
needs binary an external executable must exist on the machine
needs oauth you must complete an auth/login flow
needs node package dependency planning likely touches Node ecosystem tooling

Beginner Workflow For Skills

  1. start with built-in/native flows first
  2. enable only the skills you understand
  3. fill required env values
  4. read the skill's setup assumptions
  5. test one narrow use case

Good beginner question: - \"Do I need this skill right now, or is native Kabot functionality already enough?\"

Auto-Selected Skills

Kabot can match relevant skills to a prompt automatically.

Recent runtime improvements keep this more efficient by: - using faster matching paths for explicit skill requests - avoiding full skill-catalog bloat on lightweight prompts - preserving AI-driven responses instead of forcing stiff hardcoded behavior

This matters because the goal is: - natural conversation first - deterministic help only where confidence is high - less prompt bloat - better runtime speed

Explicit Skill Requests

You can also ask Kabot to use a skill more directly.

Examples: - please use the weather skill for this request - tolong pakai skill weather untuk ini - use the debugging workflow before changing code

Recent runtime work made these explicit skill turns much lighter, so they no longer need to drag the full skill catalog into the prompt.

Creating New Skills

Kabot can also help create or install skills through guided conversation.

Good example prompts: - create a new skill for Threads posting - install a skill from this GitHub repo - show me installable curated skills

Expected flow: 1. discovery questions 2. short plan 3. your approval 4. execution

That approval step matters.

Kabot should not jump straight from: - vague idea to - file creation and mutation

without a plan and your go-ahead.

Installing Third-Party Skills

There is a difference between: - built-in skills already shipped with Kabot - workspace skills you create - external skills pulled from another repo

For external skills, always think about: - trust - env requirements - binaries needed - who will maintain the dependency later

Common Mistakes

Mistake 1: Assuming skill enabled means fully ready

A skill may be enabled but still not operational because: - env keys are missing - external CLI is missing - OAuth is incomplete

Mistake 2: Treating skills as magic parsers

Skills improve behavior and capability, but they do not remove the need for: - clear prompts - proper setup - correct runtime environment

Mistake 3: Installing too much too early

Start with a small set of useful skills, not everything.

Good Skill Strategy

For most users:

Phase 1

  • use native Kabot first
  • add only one or two key skills

Phase 2

  • add integration skills that clearly solve a real need

Phase 3

  • create workspace-specific skills for your own workflows

Best Practices

  • enable only what you understand
  • fill in env keys before expecting a skill to work fully
  • treat third-party installs as real operational dependencies
  • keep native built-ins and external skill tooling conceptually separate