Skip to content
skillintermediate In Agent Skills Starter Kit

Claude skill packaging guide

Learn how Agent Skills are structured, draft one from your own workflow, and package it with a downloadable skill-packager skill that validates it.

You'll need: A repeatable workflow you want an agent to perform · Claude Code, or a Claude plan with custom skills enabled
  1. 1Understand the anatomy

    A skill is a folder containing a SKILL.md file and, optionally, supporting files (for example scripts/, references/ and assets/). SKILL.md starts with YAML frontmatter containing a name (lowercase letters, numbers and hyphens) and a description saying what the skill does and when to use it. The agent initially sees only the name and description of each installed skill; it reads the body when the skill is relevant and opens supporting files only when the instructions point to them. So: make the description specific enough to trigger reliably, keep the body focused on procedure, and move long reference material into separate files. Check Anthropic's current documentation for field length limits and any additional frontmatter fields.

4 steps and 2 ready-to-use prompts are in the full version

Get this resource with editable variables, examples and future updates as part of a download pack.

Get Agent Skills Starter Kit — $19

Example input

Skill name: weekly-metrics-report; reads metrics.csv, computes week-on-week deltas, writes a 200-word summary.

Example output

--- name: weekly-metrics-report description: Writes the team's weekly metrics update from reports/metrics.csv - compares last week with the week before, flags changes over 10% and produces a 200-word summary in the house format. Use when the user asks for the weekly metrics update, a metrics summary, or how the team did last week. --- # Weekly metrics report ## Goal ... ## Procedure 1. Run scripts/deltas.py reports/metrics.csv to compute week-on-week changes. ... Should NOT trigger: "build a metrics dashboard", "what is a good conversion rate?"

Troubleshooting

  • Skill never triggers: the description is probably too vague or missing the words users actually say; rewrite it with the trigger phrases.
  • Skill triggers too often: narrow the description and add 'Do not use for...' in the body.
  • Upload rejected: check the zip contains the folder (not loose files) and the frontmatter is valid YAML between two lines of three dashes.

Known limitations

  • Skill support, limits and install paths vary by product and plan and may change; verify against current documentation.
  • Skills can run code; installing one from an untrusted source is a security risk.

More from Coding & agent skills