What problem does it solve? Implementing a GitHub epic means juggling many child issues, their dependencies, and their pull requests at once. Doing it serially in one session is slow, and doing it ad hoc loses track of what is blocked, done, or stacked. This Skill turns the agent into a coordinator that fans child issues out to separate Conductor cloud workspaces, tracks each one in a ledger, links dependent PRs into GitHub stacks, and reports honest status back on the epic. ## Core Features & Use Cases - Dependency-aware fan-out: Resolves child issues from sub-issues, task lists, and back-references, maps Blocked by relationships into chains, and spawns up to 4 concurrent Conductor workspaces with contract-style prompt files. - Autonomous watching and unblocking: A single background polling loop classifies each child session via sentinel lines (EPIC-CHILD-DONE / EPIC-CHILD-BLOCKED), relays answers to how-questions, and escalates what-questions instead of inventing product decisions. - Stack management and verification: Links dependent PRs with gh stack link, cross-checks every done claim with gh pr view, detects undeclared file overlaps, and posts a final per-child report with deep links and landing instructions. - Use Case: A developer says "get epic #87 done." The Skill reads the epic, spawns Conductor workspaces for each ready child issue, grows chains as layers complete, and ends with a comment listing every PR, its stack, CI state, and any blocked or skipped children. ## Quick Start Ask the agent to work on or coordinate a specific GitHub epic by number or URL, for example: "Use the working-on-an-epic skill to drive epic #87 to reviewable PRs."
