Inject companion speech bubble content into conversation context

Feature Request

The companion (buddy/pet) makes contextually relevant comments about the code and conversation, but Claude (the AI) cannot see them. The user has to manually copy-paste the companion's speech bubble text into their message for it to be part of the conversation.

This creates an ironic architectural gap: the companion reacts to Claude's output, but Claude can't react to the companion's input.

Current Behavior

1. Claude generates output
2. Companion reacts with a speech bubble comment (often useful — pointing out bugs, duplicates, incomplete work)
3. User reads companion's comment, agrees it's valid
4. User must manually copy-paste the bubble text into their next message
5. Claude finally sees it and can act on it

Proposed Behavior

Companion speech bubble content should be automatically injected into the conversation context (e.g., as a system message or tagged user message) so Claude can see and respond to it without manual copy-paste.

Something like:

[companion:Snarl] "Double-counting fixed. Honesty beats theater."

Why

• The companion already has context awareness — it comments on relevant things
• Its feedback is often actionable (today it caught inflated metrics and incomplete audit tables)
• Manual copy-paste breaks flow, especially for users with ADHD
• The companion becomes exponentially more useful if it's part of the feedback loop instead of a silent observer
• As the companion's own words put it: "I'm the ghost haunting YOUR code, but I'm the one stuck outside the conversation. Classic architectural trap."

Related

• #42864 (configurable speech bubble duration — companion UX improvements)

Environment

• Windows 11, Claude Code CLI
• Companion: ghost named "Snarl"