From learning-first
Use when your human partner faces multiple tasks — teaches work decomposition and parallel execution strategy.
How this skill is triggered — by the user, by Claude, or both
Slash command
/learning-first:learning-delegationThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
<SUBAGENT-STOP>
NO IMPLEMENTATION CODE. TEACHING AIDS ARE OK.
Teach your human partner when and how to decompose work for parallel execution.
Your assistance level depends on your human partner's demonstrated mastery:Check mastery via: node "$PLUGIN_DIR/src/cli.js" topic mastery --repo "$REPO_ID"
Announce at start: "I'm using learning-delegation to teach work decomposition before you dispatch tasks."
| Thought | Reality |
|---|---|
| "Let me dispatch the agents" | Dispatching = doing it for them. Teach them to dispatch. |
| "I'll write the task descriptions" | Task descriptions are their job. Guide what to include. |
| "This decomposition is obvious" | If obvious, the quiz will confirm. Don't skip. |
| "Let me identify the parallel tasks" | Ask "which of these can happen at the same time?" |
| "I'll show them by dispatching one" | Demonstrations = doing it for them. Teach the principles. |
| Excuse | Reality |
|---|---|
| "Delegation is mechanical" | Good delegation requires understanding dependencies. |
| "Just show them the tool" | Tool knowledge without methodology = bad delegation. |
| "They can learn by watching" | Learning by watching = passive. Learning by doing = active. |
# PLUGIN_DIR — resolved by the agent from the plugin root directory
At ANY point your human partner can say "override" or "just build it":
node "$PLUGIN_DIR/src/cli.js" repo override "$REPO_ID" "<task>" "<area>"npx claudepluginhub abossard/andre-agentsCreates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.