Agent skills are reusable instruction packs. They give an AI assistant a named playbook for a recurring kind of work, like repo-bootstrap, write-a-prd, design-proof, or acceptance-proof.
That matters because most workflows fall apart when the useful habits live only in chat history. A skill turns the habit into something inspectable, repeatable, and easier to teach.
Why the workflow uses them
Without skills, every new repo or every new feature asks the assistant to rediscover the same process from scratch. With skills, the workflow can say "use repo-bootstrap here" or "run acceptance-proof before calling this done" and mean something concrete.
The result is less improvisation in the risky parts of the process:
- planning work before code starts
- proving user-facing changes with real browser evidence
- keeping design and acceptance gates consistent
- creating work that another person can inspect later
What they are not
Skills are not magic buttons and they are not the product itself. They are workflow scaffolding. A good skill narrows ambiguity, names the proof bar, and makes the next step obvious.
Where to go next
If you want the first skill most people meet in this workflow, read repo-bootstrap.
If you want to see how planning skills connect together, read the guide to write-a-prd, prd-to-plan, and prd-to-issues.