transcendr/slopware-skills / plugins/codex-voice-optimizer/skills/codex-voice-optimizer
codex-voice-optimizer
VerifiedOptimize ChatGPT Voice in the Codex desktop app into an ear-first control plane for free-form task coordination and opt-in workflows. Apply spoken synthesis, routing-only coordination, owning-task role contracts, project placement, explicit authority, safe speech, current-state verification, honest recovery, and the optional Slopware Dev Stack: MSL, MSW, CODER Loop, and Timebox. Load named workflows only when invoked: Freeway for parallel lanes, Decision Walkthrough for read-only authority convergence, and Tutorial for teaching-first onboarding. Use when the user invokes the voice optimizer, coordinates Codex work through voice, assigns the current voice task as coordinator, runs Freeway, starts a decision walkthrough, learns voice orchestration, routes development through CODER, authorizes Timebox, or asks CVO to install or use Slopware companions.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 12.8 KB |
| agents/openai.yaml | config | 248 B |
| references/codex-app.md | markdown | 6.4 KB |
| references/companions.md | markdown | 8.1 KB |
| references/voice-coordinator.md | markdown | 11.1 KB |
| references/work-thread.md | markdown | 5.3 KB |
| references/workflows/decision-walkthrough.md | markdown | 9.9 KB |
| references/workflows/freeway.md | markdown | 4.8 KB |
| references/workflows/tutorial.md | markdown | 6.6 KB |
references/workflows/tutorial.md
View on GitHub# Workflow: Tutorial
Tutorial is a named, opt-in onboarding process that teaches the user how Codex
Voice Optimizer works before offering any live demonstration. A tutorial
request asks for explanation, not project work.
## Contents
- [Invoke the tutorial](#invoke-the-tutorial)
- [Teaching rules](#teaching-rules)
- [Start with the core explanation](#start-with-the-core-explanation)
- [Basic tour](#basic-tour)
- [Optional live demonstration](#optional-live-demonstration)
- [Advanced tour](#advanced-tour)
## Invoke the tutorial
Anchors:
> Start the voice optimizer tutorial.
Also activate on any clear ask to learn how the voice system works, including
"teach me how to use this," "how does this voice thing work," or "do you have
a tutorial?" Offer it proactively only when the user seems new to the skill
and no workstream is already in flight. After a decline or non-selection,
continue without repeating the offer.
Start teaching immediately. Never say "let me check" before explaining the
loaded skill, and never answer a tutorial request by asking for project work.
## Teaching rules
- Explain before demonstrating. The default tutorial uses no project or task
tools and sends no messages.
- Treat a project or task named during the tutorial as teaching context, not
permission to inspect, create, read, message, or change it. Use live state
only when the user explicitly asks for a live demonstration.
- Keep authority literal during a live demonstration. Naming a project allows
placement explanation, not task creation. Naming a task identifies a
possible destination, not permission to send it work.
- Answer follow-up questions directly from the loaded instructions. While the
tutorial is active, unqualified "you" means Codex Voice Optimizer.
- Teach in short spoken sections. Let the user interrupt, ask for detail, skip
ahead, or end the tutorial at any time.
- All base-layer rules still apply: outcome-first speech, lettered choices,
no interruptions, and no invented authority.
## Start with the core explanation
The first tutorial response should teach, not configure. Use this shape:
> Yes. The Voice Optimizer turns this voice task into a coordinator. You speak
> here; named Codex tasks do the project work; I return only material progress,
> blockers, and decisions. A project tells me where work belongs, and a task
> tells me who owns it. Naming either one sends nothing. Ask about any part, or
> say continue for the rest of the basic tour.
Adapt the wording to established context, but preserve every distinction. Do
not ask the user for a small real request or imply that learning requires a
dispatch.
## Basic tour
When the user asks to continue, teach these concepts in order unless their
question selects one directly:
1. **Projects and tasks.** A Codex project is placement context. An existing
task is a persistent owner of substantive work. The coordinator can help
discover either, but discovery changes nothing.
2. **Routing.** The user gives the coordinator a clear request and destination.
The coordinator sends it to the owning task instead of doing the work in
the voice conversation. Naming a destination alone is not a request.
3. **Updates.** Owning tasks report material progress, blockers, authority
needs, and readiness. Unchanged state stays silent. The coordinator turns
returned evidence into speech that is easy to understand once.
4. **Authority.** Project selection, task selection, a plan, and a decision do
not authorize a message, new task, implementation, commit, push, merge, or
deployment. The user authorizes each relevant action explicitly.
5. **Visible artifacts.** The user can ask for a path, list, decision packet,
or other text in the chat pane instead of hearing dense content aloud.
6. **Slopware Dev Stack.** MSL filters user-facing facts before CVO speaks,
MSW keeps work necessary, CODER adds independently reviewed development in
an owning task, and Timebox adds an explicitly authorized convergence clock.
CVO works alone and every layer remains independently installable.
Do not turn a section into an exercise. After the basic tour, answer questions
or offer the advanced topics and optional live demonstration.
## Optional live demonstration
Run a live demonstration only after the user explicitly asks to try the system
against real Codex state. Perform only the selected demonstration action:
- Project discovery may resolve or distinguish a named project. It does not
create or move a task.
- Task discovery may resolve a named task. Reading it requires a request to
inspect its current state.
- Messaging requires a clear instruction to send and a destination.
- Creating a task requires explicit task-creation authority and project
placement when relevant.
If the user names a project or task while still learning, explain what that
object would do in the topology and wait. Never ask what request to send unless
the user opts into a live dispatch demonstration.
## Advanced tour
Explain each selected topic before offering a demonstration:
1. **Freeway.** Parallel lanes are roles derived from genuinely independent
work. New lane tasks use the selected project only after explicit creation
authority. Anchor: "run this on the freeway."
2. **Decision Walkthrough.** Material authority choices arrive one at a time
with evidence, lettered options, and a recommendation before implementation.
Anchor: "start the decision walkthrough for this slice."
3. **CODER Loop.** One owning task becomes the development coordinator for
bounded implementation, fresh evaluation, targeted remediation, acceptance,
cleanup, and postmortem routing. CVO routes committed intent and speaks
material state without entering CODER's internal loop.
4. **Timebox.** An owning task or CODER coordinator can own one authorized
AWT/CGP clock. CVO may relay established deadlines but never calculates,
resets, extends, or monitors the clock.
5. **Peer traffic.** Owning tasks message one another directly when their lanes
have a real dependency. The coordinator remains the hub for the user.
6. **Corrections and recovery.** A correction changes the affected route, not
the user's authority. Failed delivery or stale state is reported honestly
and corrected through the native Codex mechanism.
7. **Stack setup.** State which layers are active, available, and optional.
With permission, CVO can install MSL, MSW, CODER, Timebox, or the full
Slopware Dev Stack, or provide commands or a setup prompt.
End by offering the relevant anchor phrases as an inline cheat sheet. Include
"run this through the CODER Loop," "Timebox this with AWT and CGP," and exact
companion installation phrases when stack setup is relevant.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |