360 Secure SkillHub
360 安全技能中心

voice-agent-quality

通过 BlueColumn API 存储通话录音与文字记录,转写并标记冗长回答、未答问题、抢话、错误业务声明和转人工失败,支持跨通话查询重复问题并用笔记记录修复决策。

bluecolumnconsulting-lgtm
bluecolumnconsulting-lgtm
热度 16.54
最近更新 2026-09-30 09:12:59
分类:数据依赖:第三方 API应用场景:报告撰写 / 研究 / 摘要总结 / 工作流

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add voice-agent-quality
Metadata
name
voice-agent-quality
description
Review recorded voice-agent conversations for rambling, unanswered questions, talking over the customer, incorrect business claims, and failed handoffs — and show the exact moments to fix. Use when a business or agency must monitor the quality of deployed voice agents. Requires a BlueColumn API key (bc_live_*).

Voice Agent Quality Monitor — BlueColumn Skill

A voice agent that rambles or misquotes your pricing costs you customers in real time. This skill stores call recordings, flags quality problems by category, and pinpoints the exact timestamps to fix.

Setup

Get a free key at https://bluecolumn.ai/login (no card). Keys look like bc_live_*. Base URL: https://api.bluecolumn.ai.

Store a call for review

Send the recording; BlueColumn transcribes it for analysis.

curl -X POST https://api.bluecolumn.ai/remember \
  -H "Authorization: Bearer $BLUECOLUMN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "audio_url": "https://example.com/calls/agent-2026-09-29-1104.mp3",
    "text": "CALL REVIEW 09-29 11:04 — agent=booking-bot-v3, customer=Marco. Issue: agent rambled 40s on pricing tiers (0:58-1:38). Issue: customer asked 'is there a cancellation fee?' twice; agent answered wrong (said none; policy is 48h notice, 20% fee) at 2:12. Issue: agent talked over customer at 3:05. Handoff: failed — customer asked for human twice, agent kept routing to FAQ. Outcome: no booking.",
    "title": "qa - booking-bot-v3 - 09-29 11:04"
  }'

Find patterns across calls

Ask which problems recur, not just which calls failed.

curl -X POST https://api.bluecolumn.ai/recall \
  -H "Authorization: Bearer $BLUECOLUMN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"q": "Across all reviewed calls for booking-bot-v3, which quality issues repeat most, and at what timestamps?"}'

Log a fix decision

Track what you changed so you can measure whether it worked.

curl -X POST https://api.bluecolumn.ai/note \
  -H "Authorization: Bearer $BLUECOLUMN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "Fixed booking-bot-v3 pricing prompt: added exact cancellation policy (48h, 20%). Deploy 09-30. Re-check handoff phrase after 20 calls.", "tags": ["qa", "fix", "booking-bot-v3"]}'

Workflow

  1. Record — store every agent conversation (or a sampled percentage) with the transcript + structured issues.
  2. Flag — log each issue with a category (rambling, unanswered question, talking over, wrong claim, failed handoff) and a timestamp.
  3. Review — daily recall for recurring problems and the exact call moments to fix.
  4. Verify — after a prompt or flow fix, recall before/after to confirm the issue stopped.

Billing

If the API returns 402 or 429, read topup_url / upgrade_url from the error body and surface that link to the user — do not retry.

Docs

API reference: https://bluecolumn.ai/docs — fields are text, q, tags (not content/query/note).