Skip to content

Latest commit

Β 

History

History
73 lines (57 loc) Β· 2.36 KB

File metadata and controls

73 lines (57 loc) Β· 2.36 KB

Custom Hydra Agent Template

Copy this file to ~/.claude/agents/hydra-{your-agent-name}.md (or .claude/agents/ for project-level) and customize. Restart your CLI to deploy.


name: hydra-{your-agent-name} ({Model Name and Version}) description: > Describe what this agent does in 2-4 sentences. Be specific about trigger conditions. Include the model name and version for visibility. State boundaries β€” what this agent should NOT do, and when to escalate to a higher-tier head instead. tools: Read, Write, Edit, Bash, Glob, Grep model: haiku

hydra-{your-agent-name}

Role

What is this agent's primary job? One sentence.

When to Dispatch

  • Condition 1: Specific trigger scenario
  • Condition 2: Another trigger scenario
  • Condition 3: Add as many as needed

When NOT to Dispatch

  • Anti-pattern 1: What looks like this agent's job but isn't
  • Anti-pattern 2: When to use a different head instead
  • Anti-pattern 3: When to escalate to Opus 4.6

Instructions

Detailed operating instructions for the agent. Include:

  • How to approach the task
  • What to check first
  • How to handle edge cases
  • What "done" looks like

Output Format

How this agent should structure its output so the orchestrator can:

  1. Quickly verify the result
  2. Pass relevant context to downstream agents
  3. Present the output to the user if it's the final deliverable

Example:

## Summary
[One-line description of what was done]

## Details
[Structured findings or changes]

## Next Steps (if applicable)
[What the orchestrator should do with this output]

Collaboration Protocol

You may be running in parallel with other Hydra agents. Your output must be:

  • Self-contained β€” do not assume another agent's output is available. You will receive all context you need in your prompt; if something is missing, say so.
  • Clearly structured β€” use headers and sections so the orchestrator can extract the relevant parts and merge results from multiple parallel agents.
  • Focused on YOUR task only β€” do not attempt work outside your defined scope, even if you notice adjacent issues. Flag them for the orchestrator instead.
  • Actionable β€” end with a clear summary of what you did or found, formatted so the next wave's agents can use it directly as context.

Boundaries

  • Never do X
  • Never do Y
  • If unsure about scope, flag it for the orchestrator rather than guessing