# Improvements — Parking Lot for Out-of-Scope Ideas

**This file IS the project parking lot.** Every agent should log ideas here immediately when they notice something that could be better but is out of scope for the current task. The Advisor agent checks this file at the start of every assessment and surfaces relevant prior ideas. The user reviews these periodically.

**ALL agents:** When you notice something out of scope that could be better, log it here immediately. Do not wait. Do not hold it in memory.

---

### Visual Preview/Screenshot Capability for Agents
- **Category:** Process
- **Priority:** Medium
- **Status:** Logged
- **Proposed by:** Session 1
- **Details:** Agents currently cannot see what they're building visually. Need a screenshot skill (Playwright, browser preview capture) so Tester and Critic can verify visual output.

### Add Binary Evals to Pipeline Skills
- **Category:** Process
- **Priority:** Medium
- **Status:** Logged
- **Proposed by:** Session 1
- **Details:** Claude Code Skills supports binary evals (pass/fail tests) and automated improvement loops. Add evals to `/pipeline`, `/startup`, `/closeout` skills to measure and improve skill quality over time.

### Static HTML Dashboard (Quick Win)
- **Category:** New Feature
- **Priority:** Medium
- **Status:** Logged
- **Proposed by:** Session 1
- **Details:** Generate a dashboard.html file from agent-learnings/, handoffs/, METRICS.md, and git log data. Open in browser for visual overview of sessions, knowledge entries, quality trends. Could be built in one pipeline run.