Multi-Role Quality Synthesis: Preserve different perspectives and reach actionable quality decisions
The same release can present four different risk pictures to product, QA, engineering, and operations. Synthesis is not stapling reports together; it resolves evidence conflicts into shared decision language.
The Multi-role Quality Synthesis Skill aligns perspectives, assumptions, and unresolved risks so cross-functional teams can decide without pretending disagreement does not exist.
This guide uses collaboration-grounded examples to unpack the judgment process, helping you turn different signals into actions that are communicable and traceable.
Awesome QA Skills organizes Skills by language and testing stage. The series overview covers repository structure and shared installation options; this guide stays with Multi-Role Quality Synthesis.
Read the source Skill first
The main prompt covers Input contract, Input validation, Classification rules, Deterministic merge rules, Severity rules. Those headings are navigation; the project artifacts still provide the facts.
This Skill has no separate examples directory. The sample below is derived from its entry point, main prompt, and eval constraints.
A review fragment that keeps disagreement visible
Combine product, QA, engineering, and UX evidence without erasing disagreement
Different roles can read the same artifact and find different risks.
| Observation | Evidence | Impact | Recommendation |
|---|---|---|---|
| A critical state has no user feedback | Prototype shows only success | Users may submit twice | Add failure and recovery states |
| The schedule has no regression window | Release plan leaves half a day | High-risk fixes cannot be retested | Reduce scope or move the gate |
Do not force agreement. Record evidence and impact, then leave the decision with the accountable owner.
Merge evidence without flattening disagreement
For Combine product, QA, engineering, and UX evidence without erasing disagreement, product may care about completion rate, QA about recovery, and engineering about observability and rollback cost. Keep those observations separate.
| Role | Observation | Evidence needed | Decision ownership |
|---|---|---|---|
| Product | Can users finish the journey | Funnel and acceptance record | Scope trade-off |
| QA | Can errors and recovery be verified | Cases, runs, defects | Test conclusion |
| Engineering | Can failures be diagnosed and rolled back | Logs, metrics, deployment record | Technical risk |
Write shared facts after the role-specific notes. A conflict becomes an open decision with options and an approver. Claims about Input contract and Input validation need source links because paraphrasing can change their meaning.
The final quality note can stay short: what is proven, what remains unknown, and who owns it.
A prompt you can adapt
Replace the bracketed fields with project facts. Specific material leaves less room for guessing.
Use the multi-role-quality-synthesis Skill.
Task: Combine product, QA, engineering, and UX evidence without erasing disagreement
Version and environment: [requirement / build / environment]
Inputs: [file paths or links]
Scope: [included and excluded journeys]
Constraints: [accounts, data, time, compliance]
Keep each role observation, evidence, and impact separate. Distinguish shared facts from disagreement and name the approver for open decisions.
Finish with open questions. Do not invent missing facts.Use the first pass to inspect structure and gaps. Supply missing material before asking for the handoff-ready artifact.
Advanced use, from one call to a maintained flow
Add confidence and source_id to observations. Re-evaluate conclusions as evidence changes while retaining prior versions and decision records.
A three-Skill chain
requirements-analysis → multi-role-quality-synthesis → multi-role-quality-synthesis
| Handoff | Payload | Receiver check |
|---|---|---|
| Upstream to multi-role-quality-synthesis | Source versions, scope, risks, open questions | Multi-Role Quality Synthesis staleness and conflicts |
| multi-role-quality-synthesis to downstream | Primary artifact, evidence index, unfinished work | Multi-Role Quality Synthesis executability and owners |
| Feedback to multi-role-quality-synthesis | Runs, defects, new risks | Multi-Role Quality Synthesis baseline and regression update |
Do not paste three complete outputs into one large prompt. Give Multi-Role Quality Synthesis a structured summary and accessible source artifacts. It saves context and makes defects traceable.
Team gates
| Gate | Check | Failure action |
|---|---|---|
| multi-role-quality-synthesis input | Version, environment, owner, accessible sources | Stop Multi-Role Quality Synthesis and list gaps |
| multi-role-quality-synthesis artifact | Material claims carry basis and status | Return Multi-Role Quality Synthesis for evidence |
| multi-role-quality-synthesis execution | Command, exit status, report are reproducible | Classify infrastructure or test failure |
| multi-role-quality-synthesis decision | Residual risks have accepter and date | Do not enter the next stage |
Review Multi-Role Quality Synthesis adoption, human edit rate, unsupported claims, and failure-to-diagnosis time each sprint. Record a baseline for several cycles before setting targets.
A perspective Skill is not role-play
Do not invent what product, QA, or engineering would say. Multi-Role Quality Synthesis can only draw observations from available artifacts. When views conflict, record evidence, impact, and the decision owner.
Install and invoke
Install the individual Skill. The series overview carries the longer installation explanation.
npx skills add https://github.com/naodeng/awesome-qa-skills/tree/main/skills/en/testing-workflows/multi-role-quality-synthesis -gInvoke it with “Use the multi-role-quality-synthesis Skill,” then attach the real artifacts.
Two practical questions
Can one perspective overrule the others?
No. It adds observations. Keep disagreement and supporting evidence visible.
What if that role is absent?
Label the result as proxy analysis and list questions for the actual role. Do not fabricate approval.
What if roles reach opposite conclusions?
Keep both evidence sets, state the risk each role owns, and route the decision to the accountable person.
Can a perspective be written without data?
It can be an observation or assumption if the missing source and next evidence request stay visible.
Run Multi-Role Quality Synthesis against one real artifact and keep the input, output, and review notes. The fragments here establish structure; project evidence must still come from the project.
References
- Awesome QA Skills project:https://github.com/naodeng/awesome-qa-skills
- Awesome QA Skills series overview:https://inaodeng.com/en/blog/ai-testing/introduction_of_awesome_qa_skills/
- Awesome QA Skills: Multi-Role Quality Synthesis Skill source:https://github.com/naodeng/awesome-qa-skills/tree/main/skills/en/testing-workflows/multi-role-quality-synthesis
- Multi-Role Quality Synthesis Skill details:https://inaodeng.com/en/qaskills/multi-role-quality-synthesis/