本文へ移動
cccskills
無料GitHub で公開

setup-agent

Set up and connect a new Coval agent for evaluation. Guides through agent type selection, endpoint configuration, system prompt, and default resource attachment. Use when user says "set up an agent", "connect my agent", "create an agent", "add an agent", or "configure agent endpoint".

インストール方法を見る

含まれるファイル(2)

  • SKILL.md5.6 KB
  • references/connection-types.md5.5 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Agent Setup

Guide the user through connecting and configuring a new Coval agent for evaluation using the coval CLI. Follow the phases below in order, asking questions at each step.

If $ARGUMENTS contains an agent name or URL, use it to pre-fill the relevant field and skip that question.

Phase 0: Preflight

Step 1: Check authentication

coval whoami

If not authenticated, guide the user:

coval login

This prompts for an API key. Get one at https://app.coval.dev/settings (Organization > Manage > API Keys).

If the user doesn't have a Coval account, direct them to https://coval.dev to sign up.

Step 2: Inventory existing agents

coval agents list --format json

If agents exist, present a numbered list:

Existing agents:

  1. My Voice Bot (voice) — +12345678901
  2. Support Chat (chat) — https://api.example.com/chat

  N. Create a new agent

Ask: "Would you like to configure an existing agent or create a new one?"

If the user picks an existing agent, skip to Phase 3 (System Prompt) to update it, or Phase 4 (Default Resources) if the prompt is already set.

If no agents exist, proceed to Phase 1.

Phase 1: Agent Type Selection

Ask: "What type of AI agent do you have?"

Present the options:

  • voice — Receives inbound phone calls (your agent picks up)
  • outbound-voice — Your agent calls out (agent initiates the call)
  • chat — Text/API endpoint (OpenAI-compatible format)
  • sms — SMS-based agent
  • websocket — WebSocket connection (ws:// or wss://)

Load references/connection-types.md for type-specific guidance on required fields, validation, and common pitfalls. Use this to inform the questions in Phase 2.

Phase 2: Connection Configuration

Based on the selected type, collect the required connection details:

For voice / sms agents:

Ask: "What is your agent's phone number? (E.164 format, e.g. +12345678901)"

Validate the input:

  • Must start with +
  • Must contain only digits after the +
  • Must be 10-15 digits total (including country code)

For chat / outbound-voice / websocket agents:

Ask: "What is your agent's endpoint URL?"

Validate the input:

  • chat / outbound-voice: Must be a valid HTTP/HTTPS URL
  • websocket: Must be a valid WebSocket URL (ws:// or wss://)

Agent name:

Ask: "What would you like to name this agent?"

If $ARGUMENTS provided a name, confirm it: "I'll name this agent '<name>' — sound good?"

Phase 3: System Prompt (Optional)

Ask: "Do you have your agent's system prompt? Pasting it helps generate better test cases and metrics later."

  • If the user provides a prompt, store it as the prompt field.
  • If the user skips, move on — this can be added later.

Phase 4: Default Resources (Optional)

Inventory existing resources:

coval test-sets list --format json
coval metrics list --format json

If test sets or metrics exist, present them:

Available test sets:
  1. Customer Support Evaluation (3 cases)
  2. Edge Case Suite (5 cases)

Available metrics:
  1. Composite Evaluation
  2. Call Resolution
  3. Custom: Identity Verification

Ask: "Would you like to attach default test sets or metrics to this agent? These become the agent's default config for future runs."

If yes, collect the selected IDs and update the agent:

coval agents update <agent_id> --test-set-ids <id1>,<id2> --metric-ids <id1>,<id2>

If no resources exist, suggest creating them after setup:

  • "You can create test sets with /build-test-suite and metrics with /configure-metrics after setup."

Phase 5: Validation

For endpoint-based agents (chat, outbound-voice, websocket):

Suggest testing connectivity: "Want me to verify your endpoint is reachable?"

If yes, attempt a basic connectivity check and report the result.

For phone-based agents (voice, sms):

Confirm the E.164 format is correct and the number looks valid.

Show agent summary before creation:

Agent Summary:
  Name:     <name>
  Type:     <type>
  <Phone/Endpoint>: <value>
  Prompt:   <first 80 chars or "Not set">
  Defaults: <test sets and metrics or "None">

Ask: "Create this agent? (yes / edit)"

Create the agent:

# For voice/sms:
coval agents create --name "<name>" --type <type> --phone-number "<number>" --format json

# For chat/outbound-voice/websocket:
coval agents create --name "<name>" --type <type> --endpoint "<url>" --format json

Capture agent_id from the JSON response.

If the user provided a system prompt, update the agent:

coval agents update <agent_id> --prompt "<prompt>"

If default resources were selected, attach them:

coval agents update <agent_id> --test-set-ids <ids> --metric-ids <ids>

Verify the created agent:

coval agents get <agent_id> --format json

Phase 6: Summary + Next Steps

Present the created agent:

Agent created successfully!

  ID:       <agent_id>
  Name:     <name>
  Type:     <type>
  <Phone/Endpoint>: <value>

  View in dashboard: https://app.coval.dev/agents/<agent_id>

Suggest next steps based on what the user has set up:

  • Design a persona for testing: /design-persona
  • Build test cases for this agent: /build-test-suite
  • Configure metrics to evaluate: /configure-metrics
  • Run a full onboarding flow: /onboard

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Analyze a Coval accent testing report from runs across different speaker accents. Use when a user provides a Coval report URL, report export, run IDs, screenshots, or metric summary and wants evidence-backed next steps such as prompt changes, STT/confirmation adjustments, accent-robust routing, or expanded accent coverage.

日本語の概要は準備中です。原文の説明を表示しています。

coval-ai/coval-external-skills32026年9月16日 更新

Analyze a Coval adversarial / red-team testing report and turn it into an agent-hardening plan. Use when a user provides a Coval report URL, report export, run IDs, screenshots, or a per-scenario scorecard from an adversarial sweep and wants evidence-backed next steps such as prompt/guardrail changes, refusal hardening, verification fixes, escalation routing, or expanded attack coverage.

日本語の概要は準備中です。原文の説明を表示しています。

coval-ai/coval-external-skills32026年9月16日 更新

Analyze a Coval audio-quality testing report from runs across different voice, speaking-style, volume, interruption, and background-noise scenarios. Use when a user provides a Coval report URL, report export, run IDs, screenshots, or metric summary and wants evidence-backed next steps such as prompt changes, tool handling fixes, STT/TTS adjustments, trace setup, or expanded audio-scenario coverage.

日本語の概要は準備中です。原文の説明を表示しています。

coval-ai/coval-external-skills32026年9月16日 更新

Build or improve a Coval dashboard with metric visualizations backed by real data. Creates new dashboards from scratch or rebuilds existing ones by analyzing usage patterns, metric frequency, and data density. Use when user says "create a dashboard", "build a dashboard", "improve my dashboard", "add widgets", "visualize my metrics", "make a performance dashboard", or "dashboard for my runs".

日本語の概要は準備中です。原文の説明を表示しています。

coval-ai/coval-external-skills32026年9月16日 更新

Build a compact Coval test set from product requirements, observed conversation failures or a coverage gap. Design realistic scenarios with separate expected behaviors, provenance and controlled voice/persona variation.

日本語の概要は準備中です。原文の説明を表示しています。

coval-ai/coval-external-skills32026年9月16日 更新

Select and configure a Coval metric for a concrete product criterion, choosing transcript, audio, trace or deterministic evidence and testing the result. Use for metric creation or scoring design, not for final human calibration.

日本語の概要は準備中です。原文の説明を表示しています。

coval-ai/coval-external-skills32026年9月16日 更新

coval-ai のスキルをすべて見る

このスキルの問題を報告する