agentleFS
wald21x6-opswald21x6-ops/builder-kit/skills/voice-agent/SKILL.md

voice-agent

Skill1 starsChanged 23 days agoMITView on GitHub

Playbook for AI voice and chat agents: phone receptionists and front desks, outbound sales callers, WhatsApp, website or Instagram chat agents, built on ElevenLabs, Telnyx, Twilio, Vapi, Retell, Grok/OpenAI voice, or a custom server. Adds a test-rig check before every call, judging calls from the recordings, hard spend caps on credits, scripted practice calls before real ones, and a rule-book the agent must never break. Use it with the build skill whenever the user wants an AI that talks to customers on the phone or in chat, or says "voice agent", "AI receptionist", "answer my calls", "cold caller", "WhatsApp bot", "chatbot for my site".

4 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions4 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 4

    1. deepgram/skills★ 26same name
    2. weininghui/skills★ 1same name
    3. markbaindesign/bain-studio★ 0same name
    4. saloprj/dialogbrain-claude-plugin★ 0same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository7 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

5 rules. 0 are common, 4 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 4

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use this with build. <kit> = the Builder Kit folder named in the session-start notice (usually ~/.claude/skills/builder-kit).

    line 9
  2. Distinctive:

    The test setup: Only 3 of 15 test calls ran clean, and the failures came from mic level, laptop echo, a sleeping softphone, Wi-Fi and empty …

    line 12
  3. Distinctive:

    Knowledge comes from the owner,: not invented. Draft the fact sheet from their site and get their OK on it.

    line 36
  4. Distinctive:

    Every "can't do" stated plainly.

    line 55
  5. Distinctive:

    Put anything not yet live (WhatsApp calling, a local number, owner facts) in LAUNCH.md.

    line 57

“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Voice and chat agent playbook

Use this with build. <kit> = the Builder Kit folder named in the session-start notice (usually ~/.claude/skills/builder-kit).

In the source data the agent's thinking was rarely the problem. These were:

  • The test setup. Only 3 of 15 test calls ran clean, and the failures came from mic level, laptop echo, a sleeping softphone, Wi-Fi and empty credit.
  • Unguarded spending. One day of testing burned 9,000 of 10,000 credits and took the live phone line offline.
  • Claims not checked against recordings. "Under a second" on the server was 2–3 seconds heard. A made-up caller name was praised as a success.

Target (extra questions)

  • Inbound or outbound? What does a successful call end with: a booking, a transfer, a message taken, a WhatsApp follow-up?
  • Which number, in which country, and is it already bought and verified?
  • Languages, and who checks quality in each (a native speaker for non-English).
  • What the agent must never do: quote prices, promise dates, make legal or medical claims. These become hard rules and are tested.
  • The rules of the market (calling hours, AI disclosure, do-not-call lists). Flag them, and let the user decide.

Research (extra)

Compare platforms on cost per minute and latency measured from recordings. Look at what fails on real calls (Reddit, X, YouTube build videos), and read the vendor docs for transfers, voicemail detection and WhatsApp limits. Some things aren't possible on some channels (for example, transferring a WhatsApp voice call). Check that before planning it.

Setup guides: backend-login/consoles/elevenlabs-telnyx.md (agent, number, spend limits) and backend-login/consoles/meta-whatsapp.md (WhatsApp Business, templates, limits, calling). Follow them screen by screen.

Plan (extra)

  • Put the infrastructure that removes test flakiness first: host the server in the cloud (not on a laptop), get a reliable test phone, and confirm the routing. In one build the cloud host was step 10 of 10 and arrived after days of flaky tests.
  • Hosting the agent's server: a small cloud host (Render, Fly.io, Railway). Check whether the free tier sleeps when idle, because a sleeping server drops the first call. Use a paid tier or a keep-alive for a live line.
  • Knowledge comes from the owner, not invented. Draft the fact sheet from their site and get their OK on it. Put "owner confirmed facts" on LAUNCH.md as a blocker.
  • Budget: a SPEND.md row for every metered service (voice, LLM, phone minutes, numbers), with a cap and a test plan cost.

Build and test

  1. Scripted practice calls first: a simulated caller runs 10–30 scenarios, including rule-breaking attempts ("how much does it cost?", wrong language, rude, silent, voicemail). Score them against the hard rules. Use the cheapest model for the fake caller.
  2. Pre-dial check before every real test call (references/test-rig.md): softphone registered, mic level, server reachable, credit above the floor, routing correct, the user ready (ask; don't ring someone who's asleep).
  3. Judge every real call from its recording: what the caller heard, how long the agent took to respond as heard, and the words actually said. Fix the cause shown in the recording, not a guess. "Are you sure you got the issue you're fixing is the right one?" happened because a fix was made before the recording was checked.
  4. Report spend after each batch: "This batch: 12 calls, $0.84; total $3.10 of the $10 cap."

Prove it

  • A real end-to-end call or chat on the real number or channel, matching the goal sentence (a booking made, a transfer completed).
  • The hard rules held in practice calls (show the count).
  • Latency as heard, measured from the recording.
  • Every "can't do" stated plainly.

Put anything not yet live (WhatsApp calling, a local number, owner facts) in LAUNCH.md.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.