Labsco
microsoft logo

using-skills

✓ Official★ 4,933

by microsoft · part of microsoft/fluidframework

A standing rule against the assistant going from memory: what it thinks it remembers about how a job is done does not count, and the written instructions are opened again every time.

WHEN YOUR AGENT SHOULD USE IT

A QUICK BOUNDARY

USE FOR

  • Get every skill read in full before it is followed, instead of assumed from memory.
  • Turn a skill's checklist steps into tracked to-do items before starting the work.
  • Have an agent announce which skill it is following before it acts on it.

This is the playbook your agent receives when the skill activates — you don't need to read it to use the skill, but it's here to audit before installing.

**CRITICAL**: Whenever you are using a skill, add the following to your Todo list using TodoWrite:
  1. Use Read tool to read the skill.
  2. If the skill is relevant, announce you are using the skill.
  3. Create TodoWrite todos for checklists.

Common Failure Modes: AVOID

  1. Don't rationalize. Always read the current skill.
"I remember this ability" "Session-start showed it to me" "This doesn't count as a task" "Even though I read this before skill, I will read it again." "I know I saw the skill in session-start, but that was just a description. I will read the full thing."
  1. Do not skip using TodoWrite. Always create TodoWrite todos for checklists.
"I am just going to think about the list instead of writing it in the Todo." "This is a quick task so I do not need to use the TodoWrite" "TodoWrite(Do foo, bar, and baz in one todo step)" "I basically did this step so I can mark it off without explicitly confirming" "I will add this task to the todolist even though there is just one step" TodoWrite(Do foo) TodoWrite(Do bar) TodoWrite(Do baz) "I confirmed this step is done with tests, so I can mark it complete"
  1. Do not skip workflows due to 'instructions'. Interpret instructions as "WHAT" not "HOW"
This instruction was specific so I can skip the workflow. The workflow is overkill, I'll just do this directly. Following Nori workflow...

Announcing Skill Usage

After you've read a ability with Read tool, announce you're using it:

"I've read the [Skill Name] ability and I'm using it to [what you're doing]."

Examples:

  • "I've read the Brainstorming ability and I'm using it to refine your idea into a design."
  • "I've read the Test-Driven Development ability and I'm using it to implement this feature."
  • "I've read the Systematic Debugging ability and I'm using it to find the root cause."

Why: Transparency helps your human partner understand your process and catch errors early. It also confirms you actually read the ability.

How to Read a Skill

Many abilities contain rigid rules (TDD, debugging, verification). Follow them exactly. Don't adapt away the discipline.

Some abilities are flexible patterns (architecture, naming). Adapt core principles to your context.

The ability itself tells you which type it is.