Agent Skills
The open Agent Skills specification defines a skill as a folder containing a SKILL.md file. That file holds metadata (at minimum a name and a description) and instructions that tell the agent how to do a specific task. The folder can also hold scripts, reference documents, templates and other resources. Skills package a team's procedures into portable, version-controlled folders that any agent supporting the format can use.
Skills load in stages, which the specification calls progressive disclosure. At startup the agent reads only each skill's name and description. When a task matches a description, it reads the full SKILL.md. It opens the bundled files only when the instructions point to them. That keeps the context window small even with many skills installed, which is a context engineering win. The description decides when a skill fires, so write it as the situations and phrases that should trigger it. Qurtoo's piece on writing skill definitions that work across coding agents has more examples.
Example
---
name: monthly-close-checklist
description: Use when someone asks to close the books for a month, reconcile accounts at month end, or prepare the month-end report.
---
# Monthly close
1. Export the trial balance for the month with scripts/export_tb.py.
2. Reconcile each bank account against its statement. List any difference over $50.
3. Fill in references/close-report-template.md.
4. Stop and ask a person before posting any adjusting entry.When to use it
- A procedure you or your team repeat, with steps and gotchas worth writing down
- Specialist knowledge the agent only needs for some tasks
- You want many capabilities available without loading them all into every prompt
When not to use it
- Rules that must apply to every task; put those in the system prompt or AGENTS.md
- One-off tasks you won't repeat
- Anything secret, since skill files are read as plain text

