01Section
Getting Started Positioning
A skill is a reusable task playbook with scope, guardrails, and reference entry points.
It becomes most valuable when the same type of work happens repeatedly.
AI Programming Docs
Topic-based notes on AI programming, tooling, automation, and practical development workflows.
Choose a topic on the left.
Read the article in the center.
Use page anchors on the right.
Article
Skills
Skills turn repeated collaboration patterns into reusable capability units so agents can act on stable triggers instead of fresh prompting every time.
02Section
How It Differs from Plain Prompts
Prompts are often one-off while skills are operationalized for repeated use.
Skills define triggers, execution order, references, and fallback behavior.
03Section
What a Good Skill Must Define
A good skill explains when to use it, what to do first, what references to read, and when to stop or switch approach.
Without both actionability and boundaries, reuse becomes unreliable.
04Section
What Work Should Become a Skill
Tasks that are frequent, stepwise, and easy to validate are ideal skill candidates.
Rare tasks that depend heavily on situational judgment are usually weaker candidates.
05Section
Practical Advice
Start from one high-frequency pain point and prove the skill across several real tasks.
Skill maturity is measured by repeated reduction of rework, not by volume alone.
06Section
Common Pitfalls
A skill can fail by being too granular, too broad, or too abstract to execute consistently.
Concrete commands, checks, and output expectations are often what make it useful.
07Section
How to Build the First Example
The first skill should target one frequent, repeatable, and verifiable task.
If it reduces rework across two or three real tasks, it is worth keeping and expanding.
08Section
Recommended Learning Order
Start with single-scenario skills, then adjacent scenarios, and only later build routing or composition skills.
Abstract mega-skills look powerful but usually fail to guide execution well.
09Section
How It Combines with Other Topics
Skills combine naturally with Codex, Claude Code, OpenCode, and agent workflows.
They do not replace the model; they stabilize the collaboration method around it.