Skills — Reusable Workflows Without Prompt Repetition

Why Skills Are Helpful

Skills turn repeated prompt patterns into reusable capability blocks. Instead of rewriting “how to do this task” every session, you define it once and reuse it.

Skills are especially helpful for:

  • Consistency — same workflow every time (tests, migration checks, API review)
  • Speed — less prompt writing, faster task starts
  • Onboarding — new contributors can run team workflows immediately
  • Quality — less variance between different users and sessions

Basic Usage

At a high level, every skill includes:

  • a clear name
  • a short description of when to use it
  • step-by-step instructions the agent can execute
  • optional scripts/assets for repeatable execution

Selection is usually intent-driven: the agent picks a skill when the task matches the description, or you invoke it explicitly by name if your tool supports explicit mentions.

In mention-based tools, this is typically an @skill-name style invocation.

Claude Code Skills

Claude Code supports project and personal skills:

  • project skills: .claude/skills/
  • personal skills: ~/.claude/skills/

This model is useful for separating team-standard workflows (repo-local) from individual productivity workflows (personal).

Claude can auto-select skills from descriptions, and users can invoke specific skills directly when they want deterministic behavior.

Codex Skills

Codex supports reusable skills loaded from a skills directory and invoked when task intent matches a skill.

Use Codex skills for structured workflows that should stay consistent across sessions, especially for repository-specific checks and conventions.

Easy Start: Skill Builder Pattern

Many coding-agent setups include a skill-builder skill (or equivalent template workflow). Use it first to bootstrap your initial skill set.

Practical first three skills:

  1. test-changed-files
  2. update-docs-for-feature
  3. review-risky-diff

These give immediate value with low setup complexity.

Discovering Skills

Common Mistakes

Too broad. A skill named “do-everything” is rarely reused well.

No trigger clarity. If the description does not state when to use the skill, selection quality drops.

No maintenance owner. Treat skills like code: review, version, and update them as architecture changes.

References


Back to top

guide-LLiMes — Build LLM coding guidelines that work.

This site uses Just the Docs, a documentation theme for Jekyll.