What problem does it solve? It automates the implementation of a GitHub issue's acceptance criteria by an AI agent, ensuring every change is committed on a dedicated branch, backed by verifiable evidence, and gated by an independent reviewer before merge. ## Core Features & Use Cases - Issue-driven implementation: Reads the issue body and comments (including rework dispatch notes and prior reviewer findings) and implements only the stated acceptance criteria. - Evidence-gated workflow: Records per-AC evidence with commit SHAs, validates it locally with ztrack check, and pushes the updated body back to the GitHub issue. - Safe escalation path: Stops and posts a structured decision handoff comment instead of guessing when work requires architectural judgment or touches human-required paths. - Use Case: A maintainer comments /agent develop on issue #42; the agent implements the ACs on branch agent/issue-42, records evidence, and the substrate opens an auto-merging PR gated by CI and an independent reviewer. ## Quick Start Dispatch the develop agent on a Ready GitHub issue by setting the issue number in the ZTRACK_ISSUE environment variable so it implements the acceptance criteria and pushes a review branch.
