What problem does it solve? Raw gh output is verbose, human-formatted, and expensive in tokens for coding agents. This Skill routes GitHub operations through gh-tool, which returns terse TOON output with explicit counts, stable fields, structured errors, and suggested next commands. ## Core Features & Use Cases - GitHub domain coverage: Manage issues, pull requests, reviews, CI runs and logs, releases, labels, repository metadata, projects, secrets, variables, and search from one command grammar. - Safe mutations: Write operations support --dry-run, and stateful changes like close, merge, and cancel are idempotent, reporting already_* when retried. - Raw API access: Run authenticated GitHub API calls via gh-tool api when a dedicated subcommand does not exist. - Use Case: An agent triaging a repository runs gh-tool for a live dashboard of open issues and PRs, then uses gh-tool run view --log-failed to inspect a failing CI job without flooding the context window. ## Quick Start Ask the agent to show the current repository's open issues and pull requests using gh-tool, then drill into any item with its domain subcommand.