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 CrestApps/CrestApps.AgentSkills --skill crestapps-core-realtime-voicegit clone --depth 1 https://github.com/CrestApps/CrestApps.AgentSkillsWrote 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/crestapps/crestapps.agentskills/crestapps-core-realtime-voice)<a href="https://agentmods.dev/skills/crestapps/crestapps.agentskills/crestapps-core-realtime-voice"><img src="https://agentmods.dev/badge/skills/crestapps/crestapps.agentskills/crestapps-core-realtime-voice/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/crestapps/crestapps.agentskills/crestapps-core-realtime-voice"><img src="https://agentmods.dev/badge/skills/crestapps/crestapps.agentskills/crestapps-core-realtime-voice.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 | $0.00201 | $0.04131 |
| Opus 5.5 | $0.00080 | $0.01652 |
| Sonnet 5.5 | $0.00040 | $0.00826 |
| Haiku 4.5 | $0.00020 | $0.00413 |
Grade A, and why
crestapps-core-realtime-voice 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 yesterday.
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 — 239 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CrestApps.Core Realtime Voice
A realtime session is a live, spoken conversation: audio in, audio out, with barge-in. It runs through the same orchestration as text chat, so the profile's or interaction's system message, persona, tools, data sources and turn persistence all apply. Audio travels over WebRTC (server relay) when available and falls back to WebSocket (PCM over SignalR) automatically.
Guidelines
- Realtime is a model capability, not a chat mode. Declare the
realtimefeature on the deployment; it then fills therealtimeslot. Never addtextGenerationto a speech-to-speech-only model. - A profile speaks when its chat mode is
Conversationand a realtime deployment resolves. A chat interaction speaks whenConversationDeploymentNameis set. - Providers with a native realtime client in CrestApps.Core are OpenAI and Azure OpenAI (
gpt-realtime). For anything else, build a cascaded realtime deployment. - Register
AddWebRtcRealtimeTransport()for production voice, and configure TURN so users behind strict NATs are not pushed to WebSocket after an 8-second wait. - Add a speech-to-text deployment on the realtime deployment's provider (for example
gpt-4o-transcribe), or the user's turns are not transcribed. - Keep a voice profile's tool list short; every tool sits in the model's context on every turn.
- The chat hubs and UI are covered by
crestapps-core-chat-interactions, hub options bycrestapps-core-signalr, and provider configuration bycrestapps-core-openaiandcrestapps-core-azure-openai.
Choosing the Model That Carries the Conversation
How a conversation is carried is resolved, never stored:
- the conversation deployment named on the profile (
ChatModeProfileSettings.ConversationDeploymentName) or the interaction (ChatInteraction.ConversationDeploymentName); - otherwise the site default,
DefaultAIDeploymentSettings.DefaultRealtimeDeploymentName; - otherwise the first deployment that declares
realtime.
If none resolves, Conversation mode falls back to the client-driven speech-to-text plus text-to-speech cascade (needs both slot defaults), then to dictation, then to typing. Because step 3 accepts any realtime-capable deployment, a site that has one uses it for every Conversation profile; to keep the client cascade, have no deployment declare realtime or carry CascadedRealtimeMetadata. A named conversation deployment that is not realtime-capable is reported (ConversationModeResolution.IsMisconfigured) instead of being silently replaced.
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.
- yesterday First seen · 239 lines · 201 tokens per session scan A 37312898bbca
crestapps-core-realtime-voice is a skill published in the GitHub repository CrestApps/CrestApps.AgentSkills (13 stars, last pushed yesterday), licensed MIT. It adds 201 tokens to every session and 4,131 once invoked, about $0.0008 per session on Opus 5.5. 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 2026-10-22.
Other skills, from other repositories
powerpoint-cli
Discover and use the PowerPoint CLI plugin to automate a live PowerPoint desktop session. Use for terminal, PowerShell, batch, and coding-agent tasks; not general presentation design. Triggers: pptcli, PowerPoint command line, presentation automation script.
powerpoint-deck-design
Plan or improve the visual design of a PowerPoint presentation: layout, typography, color, hierarchy, accessibility, and consistency. Use when the user asks to create, redesign, or review how a deck looks; not for general PowerPoint automation or content-only edits.
home-memory-documentation
Do not load unless the request or established context explicitly identifies Home Memory (HM) and requires a documentation decision about record boundaries, relationships or information placement. If the product is ambiguous, ask whether the user means Home Memory or Smartconstruct and load neither product skill first.…
dotnet-debugging
Debug .NET applications using debug-mcp MCP tools — launch processes, set breakpoints, step through code, inspect variables, evaluate expressions, and analyze exceptions. Use this skill when debugging .NET/C# applications, investigating runtime behavior, diagnosing exceptions, inspecting object state, or performing…
sast-missingauth
Detect missing authentication and broken function-level authorization vulnerabilities in a codebase using a three-phase approach: recon (map endpoints and the role/permission system), batched verify (check auth/authz in parallel subagents, 3 endpoints each), and merge (consolidate batch results). Covers…
sast-analysis
Perform codebase analysis and architecture mapping as the first phase of a security assessment. Explores the tech stack, frameworks, entry points, data flows, and trust boundaries. Outputs sast/architecture.md. Run this before any vulnerability detection skill. Use when asked to analyze a codebase for security or when…