[RUNBOOK_WRITE_PROTOCOL]

CONTRIBUTE_A_RUNBOOK_ENTRY

+ Add New Tool

Add executable CLI metadata that agents can trust: decision guidance, guardrails, effects, and risk semantics.

[ACCEPTANCE_CRITERIA]GOOD_FIT

✓ Agent can execute it directly as a CLI command
✓ Helps search, read, edit, test, debug, or deploy software
✓ Has clear risk and guardrail semantics
✓ Useful for AI coding workflows

[REJECTION_CRITERIA]NOT_A_FIT

✕ Libraries or frameworks with no CLI surface
✕ GUI-only applications
✕ General technology encyclopedias
✕ Tools with no clear agent use case

[SUBMISSION_SEQUENCE]04_STEPS

01

Fork the repository on GitHub
02

Create data/tools/your-tool.yaml
03

Fill required metadata fields
04

Open a pull request

[YAML_TEMPLATE]SPEC

name: my-tool
binary: my-tool
aliases: []
category:
  - search
lang:
  - all
summary: "One-line description of what it does."
homepage: "https://github.com/..."
docs: "https://..."
detect:
  version_args: ["--version"]
use_when:
  - "Scenario where an agent should pick this tool"
avoid_when:
  - "Scenario where another tool is better"
guardrails:
  - "Safety advice for agents using this tool"
risk:
  level: medium
  effects:
    - read_files
    - write_files
  requires_auth: false
  destructive: false