What problem does it solve? AI agents operating in sandboxed environments are restricted to a subset of GitHub actions, causing commands to fail without a clear recovery path. This Skill defines a structured permission format and workflow so the agent can request exactly the access it needs from the user when a GitHub command is denied. ## Core Features & Use Cases - Structured Permission Format: Defines a command.action(resource_json) syntax covering PRs, issues, and branches with read, create, update, approve, merge, and delete actions. - gh CLI and git Governance: Establishes authoritative rules for using the gh CLI with -R ORG/REPO and git for branch operations, forbidding curl or direct API scripts. - Least-Privilege Requests: Provides 11 worked examples showing how to construct lean permission strings for tasks like creating issues, approving PRs, pushing branches, and searching commits. - Use Case: An agent tries gh pr comment 123 -R myorg/myrepo and the command fails due to sandbox restrictions. The agent constructs gh.update({"org": "myorg", "repo": "myrepo", "pr": "123"}), calls the ask_permission tool, and retries the command after the user grants access. ## Quick Start Ask the agent to comment on a pull request or push a branch, and it will request the matching GitHub permission from you if the command is blocked.
