plan-execute

name	plan-execute
description	Load a saved plan from ProjectSettings/GameDeck/plans/ and execute it step-by-step.
argument-hint	<plan-name>
user-invocable	true

When this skill is invoked:

1. Require the plan name from the user's message. If they wrote /plan-execute setup-2d-roguelike, the name is setup-2d-roguelike. If no argument was provided, use the Glob tool to list ProjectSettings/GameDeck/plans/*.md and respond:

   • With saved plans: "Usage: /plan-execute <name>. Available plans: <name1>, <name2>, ..." (names without .md).
   • Empty plans dir: "No plans saved yet. Use /save-plan to create one."

   Then stop.

2. Read the plan file with the Read tool at ProjectSettings/GameDeck/plans/<name>.md (relative to cwd, which is UNITY_PROJECT_PATH). If the file doesn't exist, fall back to the same usage hint from step 1 (Glob for the available plans, list them, stop).

3. Strip the YAML frontmatter from the loaded content — the block bounded by the first pair of --- delimiters at the top of the file. Keep everything after the closing --- as the plan body.

4. Announce execution to the user:

   • "Executing plan: <name>"

5. Execute the plan body step-by-step. Treat the body as the user's working brief and follow its structure as written:

   • If the plan has numbered steps, execute them in order.
   • If the plan has ambiguities or open questions, use AskUserQuestion to clarify before acting on assumptions.
   • Use whatever tools each step requires (Edit, Bash, MCP tools, etc.) — those come from the parent agent's permissions, not from this skill's restrictions.