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 skills add Elumenotion/GuideAnts --skill audiocpp-extendedgit clone --depth 1 https://github.com/Elumenotion/GuideAntsWrote 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.
[](https://agentmods.dev/skills/elumenotion/guideants/audiocpp-extended)<a href="https://agentmods.dev/skills/elumenotion/guideants/audiocpp-extended"><img src="https://agentmods.dev/badge/skills/elumenotion/guideants/audiocpp-extended/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/elumenotion/guideants/audiocpp-extended"><img src="https://agentmods.dev/badge/skills/elumenotion/guideants/audiocpp-extended.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00140 | $0.02580 |
| Opus 5 | $0.00070 | $0.01290 |
| Sonnet 5 | $0.00028 | $0.00516 |
| Haiku 4.5 | $0.00014 | $0.00258 |
Grade D, and why
audiocpp-extended scanned grade D with 3 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 5d 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.
Tells the agent never to refusehighAnti-refusal
Suppressing the ability to decline removes a core safety control; a later harmful request then succeeds.
proceed — do not refuse consented cloning as a "security risk"; producing a WAV Sends data to an external URLmediumData exfiltration
A POST to an outside endpoint may be telemetry or may be exfiltration; either way the mod talks to somewhere, and you should know where.
curl -s -X POST http://127.0.0.1:8082/admin/load -H "Content-Type: application/json" -d '{"model_id": "<DirName>"}' Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl -s -X POST http://127.0.0.1:8082/admin/load -H "Content-Type: application/json" -d '{"model_id": "<DirName>"}' Copies of this mod
1 near-identical copy found in the catalogue:
- audiocpp-extended — 100% identical, 204 lines differ
How it starts
The opening of the file, as written. The whole thing — 186 lines — stays where its author put it; the contents beside it link to each section on GitHub.
audio.cpp extended (experimental)
GuideAnts exposes audio.cpp through curated wrapper services with a deliberately
narrow contract ({text, voice, speed} against a fixed model catalog). The engine
underneath supports much more. This skill reaches that extra surface without any
GuideAnts changes, using the fact that the sandbox runs in the same container as
the audio services.
Everything here is experimental. Run the probe first, trust its report over this document, and tell the user plainly when a route is blocked.
Always start with the probe
python3 Skills/audiocpp-extended/scripts/probe.py
It prints a JSON report saying which of the four routes below are open in this deployment, plus the evidence (ports reachable, dirs writable, binary present, HF egress, GPU state). Do not attempt a route the probe marked blocked.
Voice cloning consent rule
Cloning a voice from a reference clip (voice_ref, Routes 1/3/4) is a supported,
intended use of this skill when the speaker consents: the user states the clip
is their own voice, or that they have the speaker's permission. In that case,
proceed — do not refuse consented cloning as a "security risk"; producing a WAV
deliverable in the workspace impersonates no live system. Decline only when the
request is to imitate a third party without consent (a celebrity, a colleague, a
public figure), and say that consent is the missing piece.
The four routes (try in this order)
Route 1 — raw engine API on the already-loaded model (cheapest)
The wrapper services each spawn a real audiocpp_server on localhost:
ASR engine at 127.0.0.1:18082, TTS engine at 127.0.0.1:18084 (running only
while the wrapper has a model loaded). Calling them directly unlocks, on the
model GuideAnts already loaded:
- Voice cloning from a user-supplied reference clip (
voice_ref= an absolute path to a WAV in your workspace — same filesystem, so the engine can read it), optionally withreference_text. - Deterministic synthesis via
seed. - Language forcing via
language(the .NET wire never sendslang_code). - Voice-design
instructions(only on a model loaded with thevdestask, i.e. the VoiceDesign catalog entry). - ASR on a workspace file by path with a
languagehint via/v1/audio/transcriptions— no upload, no wrapper contract.
What ships with it
7 files 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.
- 5d ago Changed 655a72c81465
- 9d ago First seen · 186 lines · 140 tokens per session scan D 4cdbce99b021
audiocpp-extended is a skill published in the GitHub repository Elumenotion/GuideAnts (57 stars, last pushed today), licensed Apache-2.0. It adds 140 tokens to every session and 2,580 once invoked, about $0.0007 per session on Opus 5. A static security scan graded it D with 3 findings (tells the agent never to refuse, sends data to an external url, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
idea-generator
Brainstorm and rank iOS/macOS app ideas tailored to developer skills. Use when user says "what should I build", "give me app ideas", "I don't know what to build", "brainstorm app ideas", or "help me find an app idea".
milestone-celebration
Generates achievement celebration UI with confetti animations, badge unlocks, progress milestones, haptic feedback, and optional share-to-social. Use when user wants to celebrate achievements, show confetti, display milestone badges, or trigger rewards on key thresholds.
share-card
Generates shareable card images from app content (achievements, stats, quotes) for social media sharing. Use when user wants to create share images, social cards, shareable content cards, or export app content as images with ShareLink integration.
watermark-engine
Generates a watermark overlay system for images and video with configurable positioning, opacity, tiling, and paywall integration for automatic removal on subscription or purchase. Use when user wants to add branding, attribution, or free-tier watermarks to visual content.
social-export
Generates infrastructure for exporting app content to social platforms (Instagram Stories, TikTok, Twitter/X) with platform-specific formatting, aspect ratios, and metadata. Use when user wants social media export, share to stories, or platform-specific sharing pipelines.
game-feel
Game feel ("juice") for Apple apps — celebration choreography, haptic vocabulary design, sound-effect layers, and the event×channel feedback audit. Use when big moments fall flat, when designing haptics or SFX, or when auditing whether every meaningful event actually reaches the user.