See

See

Resolve the target skill name, read its SKILL.md, and return the file contents exactly so the user can inspect the prompt text without navigating manually.

Resolve the target skill

Determine the target skill name in this order:

1. Parse the first token after the explicit see invocation.
2. If no second token is present, use the most recently created skill in the current conversation history.
3. If neither source yields a target skill, treat the request as malformed.

Normalize the target skill name before lookup:

• Strip a leading / or $.
• Trim whitespace.
• Treat the remaining text as the exact skill name.

Find the skill file

Prefer a concrete skill path already established in the current conversation history when one exists for the target skill.

Otherwise, check these exact locations in order and use the first match:

1. ~/.codex/skills/<skill-name>/SKILL.md
2. ~/.agents/skills/<skill-name>/SKILL.md
3. .agents/skills/<skill-name>/SKILL.md in the current working directory and then each parent directory up to the repository root

Do not fuzzy-match similar names. Only open an exact <skill-name>/SKILL.md match.

Output

Read the target SKILL.md file and return its contents verbatim in a fenced markdown code block.

Do not summarize, explain, or prepend navigation text. Do not paraphrase. Do not include any files other than SKILL.md unless the user explicitly asks for them.