Skills Quality assurance

Peer review design

Central QA teams measure against a standard. Peer review measures something else: whether agents share that standard, notice quality in others' work, and apply criteria when the evaluator has no audit authority.

Quality assuranceScorecards and calibrationPlaybookAny helpdeskRead-only
Installnpx rulebase-skills install cx-peer-review-design

When to use it

Reach for this when someone says any of these — they are the phrases the skill itself triggers on:

  • peer review programme
  • agents reviewing each other
  • self-QA
  • peer QA vs central QA team

How it works

The method, in the order the skill runs it. The full procedure — tables, worked examples and the edge cases — is in the skill itself.

  1. What peer review measures that audit does not

    Peer review answers: "Do we see quality the same way?" Audit answers: "Did we meet the standard?" Only the second belongs in compliance and bonus files.

  2. Step 1: Name the purpose — pick one primary

    Development (recommended default)

  3. Step 2: Design the peer loop

    Minimum viable loop: Assign, Blind where possible, Short rubric subset, Timebox, Debrief.

  4. Step 3: Self-review — when and how

    Self-review before audit or peer review can improve reflection if: Agent scores privately first, then compares to external verdict in 1:1, Self scores are not averaged into official QA, Criteria are concrete enough that self-assessment is possible.

  5. Step 4: Bias risks and controls

    Audit spot-check: central QA re-scores 10–20% of peer-reviewed conversations monthly. Large systematic gap → pause peer programme metrics, fix training.

  6. Step 5: Separate reporting lines

    Never plot peer scores on the same dashboard series as audit scores without labelling and without explaining expected offset.

  7. Step 6: When peer review substitutes audit — say no

    Peer-only QA is inappropriate when: Scores gate pay, promotion, or termination, Regulators or contracts require independent review, Dispute volume is already high, Criteria include compliance auto-fails, Team is remote/multi-site with weak trust.

  8. Step 7: Launch and evaluate

    First 90 days, track: Participation rate, Peer vs audit delta on spot-check sample, Comment quality, Agent survey.

Related skills

Free and open source, and vendor-neutral — it reads the conversations from whichever helpdesk you already run. Browse all 149 skills · connect your helpdesk over MCP · source on GitHub

Review every conversation. Act on what it finds.

AI for customer operations, built for financial services. Specialist agents chase every issue to resolution and every stalled customer to activation.

Rulebase dashboard