fix(gateway): pivot personality on the next turn, not the one after

[keniniclaw-design]

/personality wrote config.yaml and set _ephemeral_system_prompt correctly, but the model kept replying in the previous persona's voice for one more turn: with a populated transcript, prior assistant turns act as style exemplars and outweigh the system prompt.

Only visible on the second and later swaps in a process. On a cold gateway the history is empty, nothing to imitate, and the swap looks correct — which is why a restart appears to fix it.

tui_gateway._apply_personality_to_session already handles this by injecting a pivot marker, and documents the cause: "without this, LLMs tend to continue the tone established by earlier messages in the transcript". The gateway path had no equivalent.

Stage a note keyed by session_key and prepend it to the next user message, mirroring _pending_model_notes rather than writing into the transcript — history rows need display_kind bookkeeping, and an untagged marker is the class of bug that broke /retry.

Registered in _CONVERSATION_SCOPED_STATE so a staged-but-unconsumed note cannot leak across a reset or resume into a new conversation.

What does this PR do?

Related Issue

Fixes #

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 🔒 Security fix
• 📝 Documentation update
• ✅ Tests (adding or improving test coverage)
• ♻️ Refactor (no behavior change)
• 🎯 New skill (bundled or hub)

Changes Made

How to Test

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run pytest tests/ -q and all tests pass
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform:

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — or N/A
• I've updated cli-config.yaml.example if I added/changed config keys — or N/A
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — or N/A
• I've considered cross-platform impact (Windows, macOS) per the compatibility guide — or N/A
• I've updated tool descriptions/schemas if I changed tool behavior — or N/A

For New Skills

• This skill is broadly useful to most users (if bundled) — see Contributing Guide
• SKILL.md follows the standard format (frontmatter, trigger conditions, steps, pitfalls)
• No external dependencies that aren't already available (prefer stdlib, curl, existing Hermes tools)
• I've tested the skill end-to-end: hermes --toolsets skills -q "Use the X skill to do Y"

Screenshots / Logs

[teknium1]

Thanks for identifying a real gateway/TUI parity gap. Current main's gateway /personality path still only changes _ephemeral_system_prompt (gateway/slash_commands.py:2532-2549), while the TUI adds an explicit pivot marker (tui_gateway/server.py:5543-5558).

Problems

• The submitted _pending_personality_notes dictionary needs reworking for current main. Conversation-scoped state is now structural in ConversationState (gateway/session_state.py:90-128) and boundary clearing goes through GatewayRunner._clear_conversation_scope() (gateway/run.py:21849-21863).
• The note-consumption hunk predates the TurnRunner extraction; current model-note consumption is in TurnRunner.run_sync (gateway/run.py:4896-4900).
• tests/gateway/test_personality_pivot_note.py verifies staging and /new, but does not exercise the subsequent agent turn or verify one-shot model delivery and clean persisted user text.

Suggested changes

• Port the staged pivot into ConversationState and consume it in TurnRunner.run_sync beside the existing model-switch note path.
• Add an end-to-end gateway-turn regression asserting the pivot reaches the model-facing payload once without becoming a visible/persisted user message.

Automated hermes-sweeper review.

[teknium1]

Current main has moved conversation-scoped state into gateway/session_state.py:ConversationState, whose clear() is invoked by every boundary. Please port this pivot state there rather than adding another runner-level dictionary to the legacy registry.