agentmods.dev

cadence-code CLAUDE.md

cadence-code CLAUDE.md is an instructions file for coding agents from michael-L-i/cadence-code. It costs 544 tokens per session, scanned A, original, MIT.

Project instructions for Cadence Code, a Claude Code add-on that provides a local voice interface on Apple Silicon. It converts Claude’s text to speech and captures spoken input.

In plain words
What is it for?
Use it when working on starting, speaking, listening, interrupting, or ending a Cadence Code voice conversation.
Why use it?
It explains the voice conversation flow, supported controls, model handling, and project rules so changes fit the existing integration.

Instructions file

not rated 2 3d ago A scan 544 tokens original MIT
View the source file michael-L-i/cadence-code
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add instructions/michael-l-i/cadence-code/claude-md
Clone the repo
git clone --depth 1 https://github.com/michael-L-i/cadence-code

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for cadence-code CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/michael-l-i/cadence-code/claude-md.svg)](https://agentmods.dev/instructions/michael-l-i/cadence-code/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/michael-l-i/cadence-code/claude-md"><img src="https://agentmods.dev/badge/instructions/michael-l-i/cadence-code/claude-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 544 This file is loaded in full into every session.
When invoked 544 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00544 $0.00544
Opus 5 $0.00272 $0.00272
Sonnet 5 $0.00109 $0.00109
Haiku 4.5 $0.00054 $0.00054

Measured 4d ago against content hash acf03fb17b48, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

cadence-code CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

CLAUDE.md · 49 lines

How it starts

The opening of the file, as written. The whole thing — 49 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Claude Code Guide

Project Summary

cadence-code gives Claude Code a fully local voice interface on Apple Silicon. The plugin's stdio MCP process owns the selected TTS and STT models directly for an active conversation; there is no HTTP daemon or local summarization model. Speech models run through MLX Audio. MLX is not used to generate or rewrite Claude's response.

The main integration surface is /cadence-code:start-talking. It loops between choosing models on a new install, then speaking and listening through voice_start, voice_speak, voice_listen, and voice_stop. After pressing Escape, /cadence-code:jump-in silences current audio and captures added guidance through voice_interrupt. /cadence-code:wrap-up ends the conversation and releases the models. Claude Code supplies the exact text spoken by TTS. There is no passive narration.

Read AGENTS.md for the shared project map, commands, development principles, and Git workflow. Claude-specific notes below override only when they conflict.

Claude-Specific Flow

  • Claude Code installs and wires the plugin through its normal plugin mechanism; the plugin manifest starts the lightweight stdio MCP server without manual MCP setup.
  • voice_start acquires the machine-wide voice-session lock and loads TTS and STT in the background; poll voice_status until ready is true. The first call may trigger model downloads; later turns reuse the warm models.
  • commands/start-talking.md drives the conversation. Claude acknowledges longer tasks briefly and performs the actual work silently. Detailed results remain visible in Claude Code, while voice_speak receives a separately composed, concise result or question based on the current task. Summary requests should answer the requested subject, not explain the voice machinery.
  • voice_stop normally cancels active audio, drops both model providers, clears the MLX cache, and releases the session lock. Wrap Up can let a short final message finish first. Normal MCP process exit runs the same cleanup; operating-system process teardown remains the final safety net.

Read the full file on GitHub · 49 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 4d ago First seen · 49 lines · 544 tokens per session scan A acf03fb17b48

Subscribe to this mod's changes

cadence-code CLAUDE.md is an instructions file published in the GitHub repository michael-L-i/cadence-code (2 stars, last pushed 3d ago), licensed MIT. It adds 544 tokens to every session, about $0.0027 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Apple Siliconclaude-codeclaude-code-pluginmlxspeech-to-texttext-to-speech
Related

Other instructions, from other repositories

claude-mlx-tts CLAUDE.md

Instructions for aperepel/claude-mlx-tts, covering claude.md, project overview, architecture, development commands and install plugin locally for development.

aperepel/claude-mlx-tts · 764 tokens

claude-mlx-tts AGENTS.md

Instructions for aperepel/claude-mlx-tts, covering agent instructions, quick reference and landing the plane (session completion).

aperepel/claude-mlx-tts · 333 tokens

Vocello AGENTS.md

AGENTS.md instructions for PowerBeef/Vocello, covering agents.md — vocello (qwenvoice), product and authority, start and resume work, hard invariants and domain routing.

PowerBeef/Vocello · 5,217 tokens

localvoxtral AGENTS.md

AGENTS.md instructions for T0mSIlver/localvoxtral, covering localvoxtral — agent guide, build & test — read this first on a non-mac dev box, proof culture — non-negotiable, test tiers — the short version and ci / shipping.

T0mSIlver/localvoxtral · 2,917 tokens

OpenSuperMLX AGENTS.md

AGENTS.md instructions for axot/OpenSuperMLX, covering agents.md — opensupermlx, agent behavioral guidelines, 1. think before coding, 2. simplicity first and 3. surgical changes.

axot/OpenSuperMLX · 5,990 tokens

localvoxtral CLAUDE.md

Claude Code instructions for T0mSIlver/localvoxtral, a project described as: Talk to your coding agents by voice. Realtime, fully local macOS dictation that streams words as you speak and grounds LLM polishing in the exact Claude Code session under your cursor — Ghostty, iTerm2, Terminal.app, even a herdr pane. 100%…

T0mSIlver/localvoxtral · 5 tokens