Initial commit
This commit is contained in:
30
commands/use-agent.md
Normal file
30
commands/use-agent.md
Normal file
@@ -0,0 +1,30 @@
|
||||
---
|
||||
allowed-tools: Task, Read, Glob, Bash
|
||||
description: Intelligently select and use appropriate sub-agent based on task requirements
|
||||
model: claude-sonnet-4-5-20250929
|
||||
---
|
||||
|
||||
# Use Agent
|
||||
|
||||
Analyze $ARGUMENTS to determine the most appropriate sub-agent from the .claude/agents directory and use it to handle the specified task.
|
||||
|
||||
$ARGUMENTS: [task description or agent:task format]
|
||||
|
||||
## Instructions - IMPORTANT: YOU MUST FOLLOW THESE INSTRUCTIONS EXACTLY IN THIS ORDER
|
||||
|
||||
1. Run !`ls -l ~/.claude/agents` to see available sub-agents.
|
||||
2. Parse $ARGUMENTS to identify task type, domain, and requirements
|
||||
3. If sub-agent is specified → Use specified sub-agent directly
|
||||
4. If format is "youtube-url", IMPORTANT: you must immediately send task to youtube-transcript-analyzer sub-agent.
|
||||
5. Otherwise, analyze task keywords to select appropriate sub-agent from the list of available sub-agents.
|
||||
6. Use the Task tool to spawn the selected sub-agent with appropriate parameters
|
||||
|
||||
## Context
|
||||
|
||||
Available sub-agents in @~/.claude/agents/:
|
||||
|
||||
## Output
|
||||
|
||||
- Selected sub-agent name and rationale
|
||||
- Task execution through the chosen sub-agent
|
||||
- Results from the sub-agent's processing
|
||||
Reference in New Issue
Block a user