Intent linking: merge temporally-adjacent voice segments with annotations

[BraedenBDev]

Problem

Voice segments and annotations are listed separately in the output with independent timestamps. An AI agent must manually cross-reference [00:14] Circle around X with [00:16] "this needs to have actual logos" to understand they're related. This should be automatic.

Proposed Solution

When an annotation timestamp is within ~3s of a voice segment, merge them into a single intent unit in the formatted output:

Before:

## Annotations
1. [00:14] Circle around span.agency-name
## Voice Transcript
[00:16] "this needs to have actual logos over here"

After:

## Feedback
1. [00:14] Circle around span.agency-name — "this needs to have actual logos over here"
   [screenshot]

Implementation

• Formatter change in src/shared/formatter.ts
• Match annotations to voice segments by timestamp proximity (configurable window, default 3s)
• Match annotations to screenshots by annotationIndex
• Group into "feedback items" that combine element + voice + screenshot

🤖 Generated with Claude Code