hyperpowers-agents

<skill_overview> Provides a catalog of Hyperpowers agent prompts and a consistent way to launch them via Codex subagents. </skill_overview>

<rigidity_level> MEDIUM FREEDOM - Follow the selection and launch steps, but adapt prompts to the task. </rigidity_level>

<when_to_use> Use when:

• You want a dedicated subagent to run tests or commits without cluttering the main context
• You need a focused code review or test effectiveness analysis
• You need a short, bounded investigation of the codebase
• You want to offload a self-contained task and keep the main thread clean </when_to_use>

<the_process>

1. Pick the right agent

Use this mapping:

• test-runner → references/test-runner.md → agent_type worker
• code-reviewer → references/code-reviewer.md → agent_type worker
• codebase-investigator → references/codebase-investigator.md → agent_type explorer
• internet-researcher → references/internet-researcher.md → agent_type worker
• test-effectiveness-analyst → references/test-effectiveness-analyst.md → agent_type worker

2. Read the prompt file

Open the relevant reference file and reuse the exact prompt text.

3. Spawn the agent

Use spawn_agent and pass the full prompt plus the specific task.

Example:

{
  "tool": "functions.spawn_agent",
  "parameters": {
    "agent_type": "worker",
    "message": "[PASTE prompt from references/test-runner.md]\n\nTask: Run `pytest tests/` and return summary + failures only."
  }
}

4. Parallel dispatch (if needed)

If you need multiple agents at once, dispatch them in a single multi_tool_use.parallel call so they run concurrently.

5. Integrate results

Wait for all agents to complete, then integrate their summaries into the main work. Resolve conflicts manually before making final changes. </the_process>