v0.39.7

Added

• Pins can render rich HTML: pin bodies now accept full HTML documents in addition to Markdown, preserving local attachments for embedded media while keeping preview/render surfaces sandboxed.
• Pin screenshots for visual agents: termcanvas pin render <id> --json renders a pin's Markdown or HTML body to a stable PNG under .termcanvas/pin-renders/, so agents with image-reading capability can inspect visual pins directly from the generated file path.
• Independent pin HTML preview window: pin details now expose an Open preview / 独立预览 action that opens rich pin content in a resizable 1280×900 standalone window for human inspection.

Changed

• Pin render cache is cleaned during rendering: stale render snapshots and orphaned pin render directories are pruned opportunistically when a new pin render is requested.
• Agent pin instructions mention visual rendering: injected pin guidance now teaches agents when to call termcanvas pin render <id> --json and how to use the returned image path.

Fixed

• Codex hook trust registration is documented and generated: generated Codex hook setup now records the trusted command pattern and documents the registration behavior for future agents.