Roadmap / out-of-scope for v1

Things noticed during the v1 build that DO NOT belong in v1. Park them here.

v1.1 candidates

Native messaging host fork. Today the extension requires any local daemon implementing the POST /gemini contract. To make install zero-config, add a tiny x-preflight-host Node script registered as a native messaging host. Extension talks via chrome.runtime.connectNative instead of fetch(127.0.0.1). Same scoring path, one-time npm i -g x-preflight-host && x-preflight-host --install setup.
Voice training on prior posts. Pull the last 50 posts from the author's profile, derive a brief style profile, feed it into the scoring prompt so suggestions match voice. Today the prompt asks the model to mimic the draft's tone, which is weaker.
Historical scoring dashboard. Track every draft scored and the final posted version. Show trend lines, average dim scores, dim improvements. Local IndexedDB only.

Other platforms. LinkedIn, Threads, Bluesky compose boxes. Different algorithm assumptions per platform (weights in scoring.ts would need a platform key).
Image / video analysis. Feed attached media into the scoring model so visual posts get scored on the media, not just the alt text.
Hosted backend. Optional managed tier for users who don't want to run pc-agent. Brings back the BYOK API-key path that v1 deliberately killed.
Real-time scoring per keystroke. Debounced re-score as the user types. Cost concern: each Pre-flight is one Gemini call; per-keystroke would be 20-50x.

Per-tweet thread scoring (each tweet in a thread gets its own score). Brief is single-post v1.
A/B variant generator that posts both and tracks which lands. Crosses from analyzer into autopilot territory.
Chrome Web Store auto-publish. Manual submit when v1 is stable.