tool-script-calling

name	tool-script-calling
description	Practice agent shell commands, hook scripts around shell, and MCP tool calls. Use when learning or demonstrating tools/script calling, safe repo scripts, or beforeShellExecution patterns.

Tool & script calling (learning)

Short workflow for how the agent runs commands vs calls MCP tools, using this repo’s safe examples.

Steps

1. Shell (terminal) — From repo root, run the demo scripts and interpret output:

   • ./examples/scripts/demo-info.sh — one-line JSON demo.
   • ./examples/scripts/count-docs-md.sh — read-only count of docs/*.md.
   • If needed: chmod +x examples/scripts/*.sh once.

2. Hooks (optional) — Mention or configure beforeShellExecution / afterShellExecution: audit with audit.sh or block risky patterns with block-risky-shell.sh. See QUICK-REFERENCE.md and tools-and-shell-calling.md.

3. MCP (optional) — If project-docs MCP is configured, call list_example_scripts (check tool schema first) to fetch examples/scripts/README.md. That lists scripts without executing them on the server. For broader MCP habits use @mcp-patterns.

Reference

• Full pattern write-up: tools-and-shell-calling.md