What problem does it solve? Substantial project plans written in Markdown are hard to review: comments get lost in chat threads, decisions go unrecorded, and revisions lack visible diffs. This Skill authors plans under a repository's .projects directory and publishes them to the local Big Plan server so reviewers can comment, react, decide, and toggle tasks from a browser or phone. ## Core Features & Use Cases - Plan authoring and publishing: Write plans from a template with stable H2 sections, register session provenance via POST /api/session, and derive a verified Tailscale Serve or localhost review URL. - Structured review affordances: Use task checkboxes, decide/decide-multi decision cards, comparison grids, Mermaid diagrams, and GitHub-style callouts that render as interactive HTML. - Feedback round-trip: Read .comments.json sidecars, snapshot a baseline before revising, reply to questions without resolving them, and let the Send button dispatch aggregated feedback back to the authoring Claude, Codex, or Grok session. - Use Case: Draft an architecture plan, publish it, share the https URL with a reviewer on your tailnet, then revise it next session with changed blocks highlighted green and amber against their last-reviewed snapshot. ## Quick Start Ask the AI to write a big plan for your project and publish it for review, then share the verified review URL it returns.
