book-brain-visual-reader_skill

Overview

This skill provides an enhanced, visual-aware BOOK BRAIN for LYGO Havens that combines a 3-brain filesystem with LEFT/RIGHT hemisphere protocols for visual verification. It helps agents store structured text and machine-readable state while capturing and indexing visual evidence like screenshots, charts, and dashboard images. Use it when your agent has browser or image-handling capabilities and needs rigorous cross-checking between APIs, text, and visuals.

How this skill works

The skill enforces a 3-brain filesystem: Working (tmp/current), Library (memory/reference/state/logs/tools), and Outer (external sources referenced via stubs). It adds a LEFT/RIGHT brain protocol: LEFT handles structured data, indexes, and APIs; RIGHT handles visual captures, layout checks, and anomaly detection. Agents follow a three-step workflow—LEFT (expectation), RIGHT (visual capture/compare), then reconcile and log results with timestamps and references to stored screenshots.

When to use it

• Agent has browser automation, screenshot, or OCR/vision tools available.
• You need auditable verification of web UI elements vs APIs or receipts.
• You want a non-destructive filesystem with separate visual evidence storage.
• You must log source locations and visual receipts for audits or investigations.
• You need to blend timeline, state, API, text, and visual data for decisions.

Best practices

• Keep the 3-brain directories intact; never overwrite existing files—create dated/suffixed files instead.
• Run LEFT-brain checks (indexes/APIs) before capturing visuals to define expectations.
• Save only visual artifacts that back key claims; avoid hoarding screenshots.
• Add clear INDEX entries linking visuals to related reference files and receipts.
• Log reconciliation outcomes with timestamp, data point, source URL, and stored screenshot path.

Example use cases

• Verify a blockchain dashboard value: check on-chain API (LEFT), screenshot dashboard (RIGHT), reconcile and save receipt in visual/screenshots/ with index entry.
• Audit a skill listing: read structured metadata, snapshot the listing page, record discrepancies between API data and UI.
• Monitor a service health page: use LEFT to read status API, RIGHT to capture graph or red banners, log anomalies to logs/daily_health.md.
• Collect evidence for a support ticket: capture UI, attach related state and receipts, and index visuals under reference/ for later retrieval.