Skip to main content

​
Skills

A skill is a Markdown file that teaches BOR how to do a specific kind of task — a reusable playbook the AI loads on demand. Skills are how you extend BOR’s know-how without changing its code.

​
Where skills live

server/runtime/skills.js loads skills from two places:
  • Project skills — .bed-of-roses/skills/ (scoped to this project/presence).
  • Global skills — ~/.bed-of-roses/skills/ (available everywhere).
Each skill is a Markdown file (often a folder with a primary .md). The loader makes them available to the model, which invokes one with the use_skill tool when a task matches.

​
What a skill looks like

A skill is just structured Markdown: a name, a description of when to use it, and the steps/knowledge to follow. For example, a “design” skill might encode the research-first UI process; a “deploy” skill might encode your exact release steps; a “competitor-research” skill might encode a structured market-analysis brief. The description matters most: it’s how the model decides the skill is relevant. Write it so it triggers on the right requests and not the wrong ones.

​
Using a skill

The model calls: 
<use_skill name="design"/>
The runtime loads that skill’s content and returns it as a tool result; the model then follows it for the current task. Skills compose with everything else — a skill can tell the model to research with web_search, build with create_app, and run with execute_command.

​
Bundled examples

Out of the box BOR knows about skill-style capabilities such as a research-first frontend/mobile/desktop design process, marketing/competitor research, and a deep-research harness. Your own skills sit alongside these.

​
Writing a good skill

  • Lead with a precise description of when to use it (the trigger).
  • Be concrete — exact steps, exact commands, exact output shape.
  • Reference BOR’s tools the skill should use (web_search, create_app, execute_command, …).
  • Keep it focused — one job per skill. The model picks the right one by description.

​
Skills vs. memory vs. MCP

  • Memory — facts about you and durable preferences. See Memory.
  • Skills — reusable procedures for the AI to follow.
  • MCP — external tools/data sources the AI can call. See MCP.
Use a skill when you want BOR to do something a particular way every time.