JustinLaSalle

Agentic QA Copilot — Testing skill for Claude Code

Testing community

Automated call center QA analytics agent — scores every call against a weighted rubric (with emphasis on identity verification, full solution offering, and professional tone), flags low scorers for ma.

How to install Agentic QA Copilot

This entry records only its repository, not the path inside it, so there is no exact command to give. Open JustinLaSalle/Agentic-QA-Copilot and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Agentic QA Copilot does

Alternatives in Testing

Fix Issue — by metabase - Addresses GitHub issues by taking issue number as parameter, analyzing context, implementing sol 46.5k ★
Inkline — by inkline - Structures development workflow using pnpm with emphasis on TypeScript and Vue 3 Composition API 1.4k ★
Spectra App — Spec-driven development for coding agents — a desktop app, a CLI, and skills for Codex, Claude Code, Cursor, C 735 ★

README

QA Copilot

An automated call center QA analytics agent — scores every call against a weighted rubric, flags the ones that need a manager's attention, and rolls results up into a real, persistent team report. Built with the [Claude API](https://docs.claude.com).

Live demo

**[Try it here](https://claude.ai/artifact/3Ugf8DkBsXCqmB1oN7RFDH)** — note: trying the interactive part requires a free Claude account to sign in with. This one specifically also uses Anthropic's `db` persistence capability, which requires the artifact to stay within the owner's organization rather than being fully publicly shareable — a platform requirement for that feature, not a limitation of this project.

The problem

Call center QA today is a genuine bottleneck: a manager has to personally listen to a call, score it against a rubric, and deliver feedback — which means only a small sample of any agent's calls ever actually get reviewed, and the manager's time gets eaten by a task that doesn't scale. Real coaching opportunities get missed simply because there isn't time to listen to enough calls. This project automates the review step itself — not to replace a manager's judgment, but to make sure every call gets looked at, and only the ones that genuinely need a human get escalated to one.

What it does

**Generate Demo Data** — one action generates a realistic 15-call batch across 5 fictional agents with deliberately varied quality tiers (strong / average / struggling), so the report and flagged queue have something real and interesting to show without manually scoring calls one at a time.

**Scoring** — every call is scored against a 6-category weighted rubric:

Identity Verification (20%) — did the agent confirm 2+ identifiers (name + DOB, address, phone, account number) before taking sensitive action? Scored very low if verification was skipped.
Needs & Solutions Offered (20%) — did the agent offer all reasonable solutions, not just the first one?
**Professional