fix(daemon): a needsReply report resumes the parent session-only, never as a conversation post

[Poytr1]

What

Stacked on #984 (retargets to main when it merges). The niche half of the #966 scoping (the deeper transcript-row half stays open there by design): replyToSession's local branch, the relay's lineage-reply branch, and the #800 inferred reply stamp their delivery parentReport, and the #926 agent-wake inbound rendering (postAgentWakeInbound + its transcriptPostId mint) skips report deliveries.

A child's private report into a webchat-conversation parent is injected into the parent's transcript and turn exactly as before — but is no longer committed as a live conversation post fanned to the roster. Measured before the fix: every role ack and night statement of the real webchat Werewolf game ("We kill player-4 tonight." included) was room-visible (privateReportsPostedPublicly: 8–11 per game).

Unmarked agent wakes keep #926 untouched — the existing daemon-webchat pin ('an agent-initiated wake posts the INBOUND message live') passes verbatim; only report-stamped deliveries change.

Sanctioned pin flips: postedPublicly === false per reply in the night-collection cell, privateReportsPostedPublicly === 0 in both scripted Werewolf games — both previously pinned > 0 as the measured surface.

Refs #966 (remaining half documented there), #926.

🤖 Generated with Claude Code

[agentconnect-md-test]

Approved. I found no blocking issues in the trusted eaa18e9e6cb550c3dec506667159b361d6e74096 revision.

The daemon-only parentReport marker is applied consistently to local session replies and cross-daemon lineage replies, and it is checked both before minting a canonical conversation-post ID and before live inbound fan-out. Ordinary unmarked agent-initiated wakes retain the existing #926 behavior.

Verification completed:

• daemon TypeScript typecheck
• 187 relevant daemon tests across message-agent, webchat, inferred-reply, and durable-inbox behavior
• 4 scripted webchat night-collection/Werewolf tests
• ESLint and Prettier checks on all changed files

_{sent by review-bot (Codex · gpt-5.6-sol) · open in session}

[agentconnect-md-test]

Approved. I found no blocking issues in the synchronized 572c1cab62bc0036e3e4dfb5a78deb990e81dc5f revision.

I verified the exact trusted head and reviewed the full new-base-to-head patch. The daemon-only parentReport marker remains consistently applied to local session replies and cross-daemon lineage replies, and it prevents both canonical conversation-post ID minting and live inbound fan-out while leaving ordinary unmarked agent wakes unchanged. The background-wake change visible in the synchronization delta is already part of the new base and does not conflict with this patch.

Verification completed:

• daemon TypeScript typecheck
• 187 relevant daemon tests across message-agent, webchat, inferred-reply, and durable-inbox behavior
• 4 scripted webchat night-collection/Werewolf tests
• ESLint and Prettier checks on all changed files

_{sent by review-bot (Codex · gpt-5.6-sol) · open in session}