Skip to main content

Skill Packs & SKILL.md Templates

All levels

A Skill gives Claude focused expertise it loads on demand when your request matches — so you don't re-explain your needs every time. AILmanac ships a set of ready-to-use skill packs, one per kind of user. Learn the concept in depth on the Skills page.

Ready-made packs

Grab any of these from the repo's /skills folder:

PackForWhat it does
writing-content-proWriters, marketers, creatorsBrief → draft in passes → match your voice → repurpose
developer-workflowDevelopersContext-first coding, reviews, tests, clear commits
research-analystResearchers, analystsSource-grounded synthesis with citation discipline
data-analystData & spreadsheet usersSQL/formulas + plain-language reads + stat sanity checks
business-marketingFounders, SMB ownersPositioning, copy, customer comms, light planning
student-learningStudents & learnersSocratic tutoring + teach-back (learn, don't just copy)
everyday-productivityEveryoneEmail, summaries, planning, decisions

Each bakes in the house rules: give context, iterate, and verify what matters.

Install one (Claude Code)

# personal — available in all your projects
mkdir -p ~/.claude/skills && cp -r skills/writing-content-pro ~/.claude/skills/

# or per-project — shareable with your team via the repo
mkdir -p .claude/skills && cp -r skills/writing-content-pro .claude/skills/

Claude activates the skill automatically when your request matches its description. On Claude.ai / Cowork, create a skill and paste the pack's SKILL.md (availability varies by plan).

Write your own — the minimal template

A skill is a folder with a SKILL.md: YAML frontmatter + instructions. The description is the trigger — write it as "Use when…".

---
name: my-skill
description: Use when the user is doing X — be specific about the trigger so Claude knows when to load this and when not to.
---

# My Skill

Short statement of the role/goal.

## Workflow
1. Step one…
2. Step two…

## Rules
- The non-negotiables (accuracy, privacy, output format).

## Default output
What a good response looks like.

Validity checklist

  • name is kebab-case and matches the folder.
  • description says when to use it (the trigger), not just what it is.
  • Instructions are specific and bounded — one clear job.
  • No secrets, credentials, or personal/client data.
  • States how to handle uncertainty (don't fabricate; verify).