Skip to main content

read

Read skill content to stdout. This command is primarily intended for AI agents to consume skill definitions.

Synopsis​

aiskills read [skill-names] [options]

When run without arguments or flags, an interactive prompt guides you through selecting the scope, agent(s), and skill(s) to read. This is the easiest way to use this command — no flags to remember:

aiskills read
note

Running aiskills read with flags but without skill names is an error. Flags are only valid with skill names.

Non-Interactive Mode​

Provide skill name(s) along with --agent and --project/--global. All three are required:

aiskills read commit --agent claude --project              # Project, Claude
aiskills read commit --agent claude --global               # Global, Claude
aiskills read commit --agent claude --project --global     # Both scopes, Claude
aiskills read commit --agent claude,cursor --project       # Project, Claude + Cursor
aiskills read commit --agent all --project                 # Project, all agents
aiskills read commit --agent all --project --global        # Both scopes, all agents

Reading Multiple Skills​

You can read multiple skills in a single command:

aiskills read commit review-pr --agent claude --project

When reading multiple skills, a visual separator is displayed between each skill's output.

Options​

FlagShortDescription
--agent <names>-aTarget agent(s), comma-separated or all
--project-pRead from project scope (current directory)
--global-gRead from global scope (home directory)

Valid Agent Names​

universal, claude, cursor, codex, gemini, windsurf, copilot, or all

Output​

The output includes the skill's base directory (for resolving relative references to bundled resources) followed by the full SKILL.md content:

       Reading: commit
Base directory: /path/to/.claude/skills/commit

---
name: commit
description: Write conventional commit messages
---

(skill content here...)

Skill read: commit