Skills

Skills are reusable instruction documents that provide context and rules to the AI agent. Use skills to define specialized workflows, coding patterns, or domain knowledge that the agent should follow.

Using skills

Type / in the chat input bar to trigger the skill command autocomplete. Select a skill to insert a /skill-name reference into your message — it appears as a removable pill in the input bar. The reference is sent as ordinary text. The agent decides whether to load each referenced skill based on its description and what your request needs.

Skill scopes

Skills can be defined at two levels:

Scope	Location	Applies to
Global	`~/.bricks-project-desktop/skills/`	All projects
Project	`<project>/.bricks/skills/`	Current project only

When a project skill has the same name as a global skill, the project skill takes precedence.

Skill structure

Each skill is a directory with the following structure:

my-skill/
├── SKILL.md          # Required: metadata + instructions
├── agents/           # Optional: sub-agent instruction files (.md)
├── rules/            # Optional: additional rule files (.md)
└── references/       # Optional: reference files (.md)

SKILL.md format

The SKILL.md file contains YAML frontmatter followed by markdown content:

---
name: my-skill
description: A brief description of what this skill does
metadata:
  author: Your Name
  version: 1.0.0
license: MIT
allowed-tools:
  - read_file
  - write_file
  - edit_file
---

Your skill instructions go here. This content is included
in the agent's context when the skill is activated.

## Rules

- Follow these coding patterns...
- Always use this approach...

Frontmatter fields

Field	Description
`name`	Skill name (used for display and matching)
`description`	Short description shown in the skill list
`metadata`	Object with `author` and `version` fields
`license`	License type
`allowed-tools`	List of tools the agent can use when this skill is active

Built-in skills

CTOR ships with built-in skills at both global and project scope. These are available out of the box. Built-in global skills are automatically installed to ~/.bricks-project-desktop/skills/ on startup. If you delete a built-in skill, it will not be reinstalled — you can reinstall it from the skills panel. Built-in skills display a Built-in tag in the skills list.

Global skills

brainstorming

Collaborative design exploration before creative work. Guides the agent through understanding project context, asking clarifying questions, proposing approaches with trade-offs, and getting approval before implementation. After the design is finalized, the agent transitions into plan mode to write the implementation plan. Adapted from superpowers.Use when: You have a feature idea or creative task that needs design exploration before coding. Optional — skip if the task is already well-defined.

skill-creator

Guide for creating new skills via chat. Teaches the skill creation lifecycle from planning to iteration — covering skill anatomy, progressive disclosure design, and bundled resources.Includes helper scripts for initializing, validating, and packaging skills.Use when: You want to create a new skill for the agent.

imagegen

Generates and edits raster images for the current project using CTOR’s configured image providers (OpenAI gpt-image-2, OpenAI Codex backend, or Google Nano Banana through Gemini API or Vertex AI). Includes a bundled Bun script that resolves provider credentials, writes outputs into the project (default: output/imagegen/output.png), and supports size, quality, aspect-ratio, and format options.The skill is active only when at least one image-capable provider is configured. With --provider auto, CTOR prefers the provider of the active chat model when it supports image generation, and otherwise falls back to the configured default.Use when: You need a generated bitmap asset (photo, illustration, texture, sprite, mockup, product image, UI concept) or want to edit an existing raster image. Skip when SVG, HTML/CSS, canvas, or an existing native asset is the better deterministic output.

bricks-cli

Reference for the BRICKS CLI tool (bricks-cli). Covers BRICKS Foundation management, application and module operations, media flow, LAN device discovery, ACP bridge, and MCP server integration.Key capabilities:

BRICKS Foundation management — list, control, screenshot, and monitor devices
BRICKS Foundation groups — dispatch commands, refresh, and monitor groups
Applications — list, update, bind devices, init and deploy projects
Modules — list, update, init projects
Media flow — manage media boxes, upload files with processing options
DevTools — discover devices on your local network via UDP broadcast
MCP server — start a local MCP server for tool integration
ACP bridge — bridge with the CTOR agent

Use when: You need to interact with the BRICKS API or manage BRICKS Foundation from the command line.

Extracting a reusable prompt template from an existing application is no longer a built-in skill. Install the App Templates (app-templates) plugin from the built-in marketplace instead — it updates without an app release.

Project skills

BRICKS projects include built-in skills with domain knowledge for building BRICKS applications:

bricks-ctor — architecture patterns, animations, automations, data calculations, and more
bricks-design — creative design guidance for building distinctive, polished BRICKS interfaces

bricks-ctor built-in skills

See the full skill reference in the bricks-ctor docs.

Installing skills

Install third-party skills from a public GitHub repository or a local folder. In the skills panel, click + Install Skills to open the installer.

From a GitHub repository

Pick GitHub Repository and enter a repo spec:

owner/name — uses the repository’s default branch
owner/name@branch — pin a specific branch
A full GitHub URL — https://github.com/owner/name or https://github.com/owner/name/tree/branch

Click Detect. CTOR downloads a tarball of the target branch and recursively scans it for SKILL.md files.

From a local folder

Pick Local Folder, click Pick Folder to choose a directory, then click Detect. CTOR scans the folder for SKILL.md files. If the picked folder is itself a skill (contains SKILL.md at its root), it is treated as a single skill. On macOS, the Open Folder in CTOR Finder Service shortcuts this flow. Right-click a folder in Finder and choose Services → Open Folder in CTOR: if the folder contains SKILL.md files but isn’t a BRICKS project, CTOR opens this installer pre-filled with the folder under Global scope and lands you directly on the detected-skills list.

Selecting and installing

Detected skills are listed with their names and descriptions. Toggle the ones you want to install, or use Select all / Select none. Then choose the install scope in the footer:

Global — installs to ~/.bricks-project-desktop/skills/
Project — installs to <project>/.bricks/skills/ (only available when a project is open)

Click Install (N) to install the selected skills. If any of them already exist at the target scope, CTOR shows a confirmation banner listing the conflicts and asks you to confirm before overwriting.

Managing skills

Open the Plugins panel from the sidebar and switch to the Skills tab. From here you can:

Toggle between Project and Global scope
View all available skills with their descriptions
Enable or disable a skill with the switch on each card
Edit a skill’s SKILL.md content
Delete a skill

Skills can also be installed by plugins. A plugin-installed skill carries a plugin: {name} badge, and your edits to it are kept — updates never overwrite your changes.

Enabling and disabling skills

Each skill card has a switch that toggles whether the skill is loaded into the agent’s context. Disabled skills stay listed but are skipped when you trigger / autocomplete and won’t be passed to the agent. This applies to built-in skills too — use the switch to silence a built-in skill (such as imagegen) without removing it.

Rules and references

The rules/ and references/ subdirectories can contain additional markdown files that are loaded alongside the main SKILL.md content:

Agents — instruction files for sub-agents. The main agent can pass these to spawn_agent via the instructions_file parameter to drive a sub-agent (typically general) with skill-specific instructions.
Rules — additional instructions or constraints for the agent
References — background knowledge, API docs, or examples

All .md files in the rules/ and references/ directories are included when the skill is activated. Files in agents/ are loaded on demand when referenced by instructions_file.

Getting started

Working with projects

Ship your app

Reference

Changelog

Using skills

Skill scopes

Skill structure

SKILL.md format

Frontmatter fields

Built-in skills

Global skills

Project skills

bricks-ctor built-in skills

Installing skills

From a GitHub repository

From a local folder

Selecting and installing

Managing skills

Enabling and disabling skills

Rules and references

​Using skills

​Skill scopes

​Skill structure

​SKILL.md format

​Frontmatter fields

​Built-in skills

​Global skills

​Project skills

bricks-ctor built-in skills

​Installing skills

​From a GitHub repository

​From a local folder

​Selecting and installing

​Managing skills

​Enabling and disabling skills

​Rules and references

Using skills

Skill scopes

Skill structure

SKILL.md format

Frontmatter fields

Built-in skills

Global skills

Project skills

Installing skills

From a GitHub repository

From a local folder

Selecting and installing

Managing skills

Enabling and disabling skills

Rules and references