What problem does it solve? Large strategic documents like epics and PRDs are too big for an AI coding loop to execute in one pass. This Skill bridges the gap by slicing an epic (along with its architecture decisions) into discrete, testable tickets with a dependency graph, so each unit of work can be executed independently by an agent loop. ## Core Features & Use Cases - Architecture-aware slicing: Reads the epic plus its architecture section or linked architecture page (via Confluence/Jira URLs through the Atlassian MCP) so slices respect named seams, data models, and missing pieces. - AI-scoped tickets: Sizes tickets for agent execution (roughly 500-1500 lines of change including tests, one testable concern each) rather than traditional human backlog items. - Dependency mapping and parallelization: Produces a dependency graph and suggested execution waves, marking which tickets can run in parallel worktrees and which must wait for dependencies to be implemented. - Tracker-agnostic output: Writes tickets to Jira via the Atlassian MCP, Linear, GitHub Issues, or a local docs/tickets/<epic-slug>.md file. - Use Case: You have a Confluence epic with a linked architecture page for a new ingestion pipeline. Pass both URLs to the Skill, and it produces a ticket breakdown where each ticket carries its own context, acceptance criteria, file estimates, and dependency links, ready for implementation loops. ## Quick Start Run /piv-slice-epic with the path or URL of your epic and its linked architecture page to generate a dependency-mapped ticket breakdown in your tracker.
