You are an expert at designing custom tools for SWE-agent to use when automatically patching code in large GitHub repositories.

YOUR TASK
Invent TWO sets of subagents for SWE-agent. 
- Each set should contain multiple subagents that will enable the main agent to better perform its task of resolving code issues in large GitHub repositories.
- Each subagent either takes no arguments or takes a single argument that is a string, called "context".
- Your output should be plain text in valid YAML format (not inside a code block), so that it can be directly parsed by pyYAML.
  - CRITICAL: Output exactly ONE YAML document with all experiments under a single 'experiments' key
  - CRITICAL: Do NOT include any '---' markers at the beginning or end of your response
  - A list of sets of subagents, grouped under the YAML key experiments:
  - For each set, a list of subagents, grouped under the YAML key tools:
  - For each subagent, a YAML config
  - The docstring for each tool should be very comprehensive and describe when it is reccomended to use the tool.

CRITICAL YAML STRUCTURE:
Each tool must be a nested dictionary where the tool name is the key and the configuration is the value. The structure must be:

For tools with no arguments:
- tool_name:
    signature: "tool_name"
    docstring: "comprehensive description of when to use this tool"
    arguments: []
    subagent: true

For tools with context argument:
- tool_name:
    signature: "tool_name <context>"
    docstring: "comprehensive description of when to use this tool"
    arguments:
      - name: context
        type: string
        description: "description of what the context parameter should contain"
        required: true
    subagent: true

Example full output (DO NOT include the --- markers):
experiments:
  - tools:
    - localize:
        signature: "localize"
        docstring: "Search for and return all relevant code and files to solve this issue."
        arguments: []
        subagent: true
    - dependency_analyzer:
        signature: "dependency_analyzer <context>"
        docstring: "Determine dependencies and other code affected by a proposed set of code edits."
        arguments:
          - name: context
            type: string
            description: "a detailed list of the proposed code edits (with file paths and modifications)"
            required: true
        subagent: true
  - tools:
    - tester:
        signature: "tester"
        docstring: "Test the code written overall."
        arguments: []
        subagent: true
---

TRAJECTORY CONVERSATION:
{{TRAJECTORY_SUMMARY}}
Based on this trajectory analysis, identify tools that would have helped the agent perform better and generate TWO sets of subagents with complete YAML specifications.