Batch Skill
Execute multiple tool calls in parallel within a single response.
How It Works
This is a documentation-only skill — there is no script to run. Instead, the agent should emit multiple tool_use blocks in a single response. The agent loop already handles parallel tool execution natively.
When to Use
Use batch execution when:
- Independent operations: Multiple file reads, searches, or edits that don't depend on each other
- Gathering information: Reading several files at once to understand a codebase
- Parallel writes: Writing multiple independent files simultaneously
- Speed optimization: When operations can run concurrently
How to Use
Simply include multiple tool calls in a single response. For example, to read three files at once:
[tool_use: read_file with path="src/a.ts"] [tool_use: read_file with path="src/b.ts"] [tool_use: read_file with path="src/c.ts"]
All three will execute in parallel and results will be returned together.
Guidelines
- Only batch independent operations — if tool B depends on tool A's result, run them sequentially
- The agent loop handles execution, error handling, and result aggregation automatically
- Each tool call in the batch is executed independently — one failure does not affect others
- Results are returned in the order the tools complete
