What problem does it solve? When an autonomous agent organization hits work it is not authorized to complete — a sensitive PR, an out-of-scope decision, or a missing clarification — it needs a defined way to hand that work to a person and wait for an explicit resolution. This Skill is the task spec handed to the human maintainer, explaining what is being asked and how to resolve it so the automated flow resumes. ## Core Features & Use Cases - Human-approval gating: Explains how to authorize a human-required PR via a native GitHub Approve or an exact-SHA /agent approve <full-head-sha> comment, which the human-approval gate verifies per commit. - Typed decision handling: Covers the four engagement types — approve, decide, answer, and inform — with the exact /agent decide and /agent answer commands that record results and clear blocked states. - Use Case: A PR touches workflow files listed in policy.risk.human_required_paths, so the org assigns it to you and parks the merge. You review the diff, then comment /agent approve <current-head-sha>; the gate verifies your authorization and the PR proceeds once ci, security, and agent-review are also green. ## Quick Start Open your GitHub worklist with the query is:open assignee:@me label:human-required, review the assigned item, and resolve it with a native Approve or the appropriate /agent command described in this Skill.
