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.
npx agentmods add skills/kunitoki/sonic-skills/audio-performance-debugnpx skills add kunitoki/sonic-skills --skill audio-performance-debuggit clone --depth 1 https://github.com/kunitoki/sonic-skillsWhat 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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00099 | $0.01037 |
| Opus 5 | $0.00049 | $0.00518 |
| Sonnet 5 | $0.00020 | $0.00207 |
| Haiku 4.5 | $0.00010 | $0.00104 |
Grade A, and why
audio-performance-debug 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 2d 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.
How it starts
The opening of the file, as written. The whole thing — 63 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Audio Performance Debug
Audio performance requires good worst-case complexity, not average-case — the audio thread fires on a hard deadline every buffer, regardless of what else is happening.
Step 1 — Identify the symptom
Measure before guessing. Different symptoms point to different root causes.
| Symptom | Measurement approach |
|---|---|
| Consistent high CPU average | CPU meter in DAW; perf stat / Instruments Time Profiler |
| Sporadic xruns / dropouts | Host xrun log; correlate with system events (GC, network, page fault) |
| CPU spikes on note-on | Profile with many simultaneous note-on events; watch for allocation spikes |
| Latency reporting wrong | Log getLatencySamples() before and after prepareToPlay at varying buffer sizes |
| Gets worse with more voices | Profile with 1 vs 8 vs 32 voices; O(N²) shows quadratic growth |
| Memory bandwidth pressure | perf mem / VTune memory bandwidth counter; cache miss rate |
Step 2 — Locate the hotspot
- Build a release (optimized) binary before profiling — debug builds are not representative.
- Use a reproducible test case: fixed buffer size, fixed voice count, looped audio.
- Attach a profiler to the audio thread specifically, not the whole process.
- Identify the top-3 hottest functions by exclusive CPU time, not inclusive.
- Check whether the spike is periodic (every N buffers → container rehash or GC) or random (OS jitter, page fault).
| Platform | Profiler | Notes |
|---|---|---|
| macOS | Instruments — Time Profiler | Filter to audio I/O thread; use "hide system libraries" to focus on your code |
| macOS | Instruments — Allocations | Catch allocations on the audio thread during a session |
| Linux | perf record -g + perf report |
--call-graph dwarf for C++ templates; perf stat for cache miss ratio |
| Windows | VTune Profiler — Hotspots | Use "Platform Profiler" preset; filter to realtime thread |
| Cross-platform | Tracy | Frame-level instrumentation; zero-cost when disabled; shows per-buffer timing |
| JUCE | juce::PerformanceCounter |
Inline timer around suspect blocks; logs to console |
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 2d ago First seen · 63 lines · 99 tokens per session scan A 259cd476f47f
audio-performance-debug is a skill published in the GitHub repository kunitoki/sonic-skills (18 stars, last pushed 3mo ago), licensed Unlicense. It adds 99 tokens to every session and 1,037 once invoked, about $0.0005 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 2026-08-30.
Other skills, from other repositories
AudioMind
Tired of juggling multiple audio APIs? This skill gives you one-command access to TTS, music generation, sound effects, and voice cloning. Use when you want to generate any audio without managing multiple API keys.
lattice-init
Guided setup and upgrade-check experience for Lattice projects -- scans the repository, detects existing configuration and outdated conventions, suggests refiners and available upgrades in priority order, and creates or reconciles the .lattice/ config. Bridges the gap between installing skills and getting first value…
intelligent-investor-graham
Use Graham value investing for Is this investment or speculation, defensive stock screens, margin of safety, Mr. Market, allocation, funds, IPOs.
skill-to-workflow
将现有 Skill、Prompt、SOP 或 Agent 操作说明拆解为可执行、可监控、可重试的 Workflow 规格；识别真实节点、共享子工作流、工具调用、状态、分支、循环、 提示词、数据契约和验收方式。适用于用户要求把一个 Skill 转成工作流、节点图、 编排方案或 Workflow 实现规格时；不替代原 Skill 执行业务任务，也不在未要求时 绑定具体工作流平台或直接部署。.
query-loop-implementation
Implement a production-ready LLM query loop / agent loop for AI applications. Use this skill whenever the user wants to add tool calling, ReAct-style reasoning-action-observation cycles, function calling loops, query engines, agent runtimes, toolresult feedback, max-turn exits, or Claude Code-like Agent Loop behavior…
documentation-and-adrs
Records decisions and documentation. Use when making architectural decisions, changing public APIs, shipping features, or when you need to record context that future engineers and agents will need to understand the codebase.