CCCC v0.4.12 Release Notes

v0.4.12 builds on v0.4.11 with a broad product-quality pass across Voice Secretary, built-in assistants, durable delegation, headless delivery, Runtime Dock visibility, and the Web workspace.

This release is the first version where Voice Secretary becomes a full repository-backed working surface instead of only a capture/transcript experiment. It also adds stronger task-backed collaboration primitives, clearer assistant controls, more stable headless message delivery, and a more polished Web operator experience.

Highlights

1) Voice Secretary is now a real working-document assistant

Voice Secretary received the largest upgrade in this release.

The assistant now has a dedicated workspace, repository-backed markdown documents, request history, transcript feedback, Ask/Document/Prompt modes, and a cleaner input path for document maintenance. Stable transcript input is no longer treated as a raw chat message stream; it is routed through the Voice Secretary document/request surfaces so the assistant can pull the relevant input, update the target document, and report through the correct channel.

Key improvements:

• repository-backed Voice Secretary documents under docs/voice-secretary/
• a larger Voice Secretary workspace for reading, editing, creating, archiving, and downloading markdown documents
• Document, Ask, and Prompt modes in the chat composer assistant controls
• prompt-refine flow that returns composer-ready text instead of visible chat noise
• transcript input aggregation, previous-input tail context, idle review, and stop-flush handling for better long-form document quality
• request/activity history in the Voice Secretary workspace with clearer live transcript feedback
• source and artifact reporting improvements for Ask-style answers
• less noisy Voice Secretary notifications and a narrower dedicated help surface

The default Voice Secretary behavior is also more explicit: document work edits repo markdown directly, Ask work returns a user-visible reply through the request channel, and Prompt work submits a draft back to the composer.

2) Built-in assistants have a clearer Web control surface

v0.4.12 introduces a more coherent built-in assistant control path in the chat composer and settings.

PET and Voice Secretary are now exposed as compact assistant controls near the composer rather than as disconnected settings-only features. The settings side was also reorganized around built-in assistant configuration, including larger prompt editing areas and cleaner assistant-specific settings.

Key improvements:

• chat composer assistant rail for PET and Voice Secretary
• Voice Secretary mode control for Document, Ask, and Prompt flows
• clearer assistant enable/open/record affordances
• assistant settings moved into a dedicated settings area
• larger assistant prompt editors with expand behavior
• runtime visibility controls generalized from PET-only to built-in assistant surfaces

This keeps built-in assistants visible without turning the main composer into a heavy workflow form.

3) Durable delegation is now more visible and safer to use

This release adds a stronger foundation for semi-structured collaboration without replacing the chat-first workflow.

tracked_send is now a daemon-owned delegation primitive that can create a task and linked message together, while preserving idempotency across retry paths. Task references in chat are also rendered as live chips with status projection, so operators can see whether a delegation is planned, active, waiting, handed off, or done directly from the conversation.

Key improvements:

• daemon-side tracked delegation with task/message linkage
• idempotency hardening for partial-failure retry cases
• task references rendered in message bubbles
• task chip status projection instead of static labels
• linked message context in task surfaces
• help guidance for claim-back, completion evidence, and reply routing
• cleaner inbound rendering for PTY/headless runtime messages

The result is a lighter path from natural chat to durable responsibility tracking, without adding a new heavy PM workflow.

4) Headless delivery and reply routing are more consistent

Headless Codex and Claude flows continue to be tightened in v0.4.12.

The release improves headless delivery formatting, recipient inference, source/target context, and capability search pagination. It also keeps visible replies on the canonical MCP message path instead of relying on special-case shortcuts.

Key improvements:

• better headless inbound message rendering
• safer headless reply-recipient routing for messages with ambiguous sender context
• unified delivery planning across PTY and headless actors
• cleaner final-output handling for headless runtimes
• specialized MCP tools modeled explicitly in capability surface tests
• capability search pagination improvements

These changes reduce cases where a headless actor replies to the wrong target or cannot tell who the incoming work is from.

5) Runtime Dock and workspace state are easier to trust

Runtime visibility received another polish pass after the v0.4.11 ring-state refactor.

The Runtime Dock now distinguishes idle and stopped actors more clearly, applies ring styling more consistently, and handles assistant/runtime visibility more predictably. PTY and headless actors share a clearer state vocabulary while keeping their transport-specific details behind the same visual surface.

Key improvements:

• more consistent dock ring sizing and state tone
• clearer idle versus stopped visual treatment
• PTY actors no longer look busy just because their terminal is alive
• assistant runtimes can be surfaced in runtime visibility controls
• group/runtime state refresh behavior is less likely to drift after restarts

This makes the dock a more reliable operator signal instead of a decorative status light.

6) Web UI polish and workspace chrome were tightened

The Web workspace received a large amount of product-surface cleanup.

The chat composer, assistant controls, modal layers, settings navigation, group/sidebar surfaces, copy behavior, markdown rendering, and presentation/workspace surfaces were all refined. The release also introduces shared UI primitives for buttons, dialogs, popovers, inputs, surfaces, and command-style controls.

Key improvements:

• refreshed chat composer layout and workspace chrome
• less fragile modal layering and portal handling
• improved copy fallback behavior and shared clipboard utility
• markdown rendering refinements
• grouped settings navigation and visual polish
• updated logos and packaged Web assets
• user message bubble background restored after text-color unification
• actor secret placeholders clarified for Codex and Claude Code runtimes

The Web surface should feel more coherent and less like a collection of unrelated panels.

7) Self-evolving skills and capabilities are easier to manage

The self-evolving skill workflow was refined across capability management and group-level visibility.

This release focuses on making generated skills more understandable and manageable, while keeping the mechanism closer to the existing capability model instead of creating a separate hidden system.

Key improvements:

• improved Self-Evolving Skills management UI
• group-aware self-evolving skill organization
• clearer actor assignment state
• provenance display for generated skill records
• cleaner capsule preview and readiness signals
• capability search and enablement behavior aligned for self-proposed skills

This keeps self-generated skills visible to the operator while preserving the agent-facing capsule workflow.

8) Automation follow-ups are clearer

Built-in automation settings were renamed and reorganized around collaboration follow-ups.

The Web settings no longer emphasize internal "built-in automation" terminology as strongly. Instead, the controls are grouped around what operators actually tune: message follow-ups, progress follow-ups, context refresh, and advanced foreman alerts.

Key improvements:

• clearer Collaboration Follow-ups settings copy
• reset-to-defaults behavior aligned with other settings panels
• zero-value helper text corrected to match daemon behavior
• rule/snippet settings layout polished

This improves understanding without changing the core automation engine into a heavier workflow system.

9) IM, CLI, and docs received targeted fixes

v0.4.12 includes smaller but important maintenance improvements outside the main Web assistant work.

Key improvements:

• cccc update command with install-source detection
• WeCom response URL fallback for streaming and media replies
• DingTalk mention handling and sender identity coverage
• README and public docs refreshed to match the current product direction
• standards documentation updated for new daemon IPC and assistant surfaces
• logo assets refreshed across docs and Web packages

Summary

In short, v0.4.12 is the release that brings together:

1. a full Voice Secretary workspace with repo-backed markdown documents
2. clearer built-in assistant controls for PET and Voice Secretary
3. daemon-owned tracked delegation and visible task chips
4. more consistent headless delivery and reply routing
5. more trustworthy Runtime Dock state projection
6. substantial Web UI and settings polish
7. refined self-evolving skill management
8. clearer collaboration follow-up settings
9. targeted IM, CLI, documentation, and test coverage improvements

This release also intentionally backs away from automatically generating Codex project .codex/config.toml files from CCCC actor private environment settings. Codex provider customization remains explicit through Codex's own login/configuration and command-line override mechanisms.