read

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

Synopsis

aiskills read [skill-names] [options]

Interactive Mode (Recommended)

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.

Tip

In any multi-select prompt, press Shift+Tab to toggle select/deselect all.

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

Flag	Short	Description
--agent <names>	-a	Target agent(s), comma-separated or all
--project	-p	Read from project scope (current directory)
--global	-g	Read from global scope (home directory)

Valid Agent Names

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

Output

The output starts with metadata about the skill, followed by the full SKILL.md content. Paths under your home directory are abbreviated with ~/.

         Reading: commit
  Base directory: ~/git/username/path/to/project/.claude/skills/commit
      sourceType: git
          source: owner/commit-skills
         subpath: skills/commit
          commit: 3f9c2a1d8e7b6c5a4f3e2d1c0b9a8f7e6d5c4b3a
      sourceHash: 5d6e7f8a9b0c1d2e3f4a5b6c7d8e9f0a1b2c3d4e
            name: commit

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

(skill content here...)

Skill read: commit

Output Field Reference

Field	Meaning
Reading	The skill name as installed (its directory name)
Base directory	Path to the skill on disk (uses ~/ for paths under your home, and is annotated with (from $ENV_VAR) when the agent's global directory is relocated)
sourceType	Origin type: git or local
source	Origin reference (e.g. owner/repo for git, a path for local)
subpath	Path within the source where SKILL.md lives. A skill at the repository root shows <root>
commit	The Git commit the installed content came from (Git sources only)
sourceHash	The recorded version of the skill: the Git tree hash of its source folder. See Version Check
checkedAt	When update last checked the skill (global skills only)
name	The name field from SKILL.md (may differ from the folder name)

commit, sourceHash and checkedAt are shown only when they are recorded. Skills installed by 2.10.0 or earlier show them after their next update.