qa-review
Skill37 starsChanged 36 days agoMITView on GitHub
Use when reviewing a feature for release readiness. Covers acceptance-criteria verification, edge-case enumeration, regression risk assessment, and a release decision that is a judgment rather than a vibe. Trigger on "is this ready to ship", "release sign-off", "what edge cases are we missing", "regression risk of this change". Do not use for line-by-line PR review — see the code-review skill; for the overall test plan, see the test-strategy skill.
12 other projects use a version of this.
Adopted by
Most recently taken up by
- bcvk/support-lead-kittheir own version, first seen Nov 15, 2026
- drosam/agentstheir own version, first seen Nov 15, 2026
- fushinoryuu/claude_skillstheir own version, first seen Nov 1, 2026
- drvoss/everything-copilot-clitheir own version, first seen Oct 26, 2026
- TzolkinB/skillstheir own version, first seen Oct 22, 2026
- ieuanign/skillstheir own version, first seen Oct 13, 2026
Other versions12 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 12
- nearai/ironclaw★ 13ksame name
- arjun988/blender-skills★ 2.2ksame name
- hellowind777/helloagents★ 704same name
- PackmindHub/packmind★ 323same name
- shuotao/REVIT_MCP_study★ 110same name
- drvoss/everything-copilot-cli★ 49same name
- ieuanign/skills★ 2same name
- TzolkinB/skills★ 2same name
- drosam/agents★ 1same name
- bcvk/support-lead-kit★ 0same name
2 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository137 other files it gives its agents.
- agent-instructionsSkill
- agent-memorySkill
- hooksSkill
- plugin-developmentSkill
- skill-authoringSkill
- skill-reviewSkill
- slash-commandsSkill
- subagentsSkill
- agent-designSkill
- context-engineeringSkill
- fine-tuningSkill
- llm-cost-optimizationSkill
- llm-evaluationSkill
- llm-integrationSkill
- mcp-serverSkill
- ml-pipelineSkill
- model-selectionSkill
- prompt-engineeringSkill
- ragSkill
- structured-outputSkill
- api-designSkill
- background-jobsSkill
- cachingSkill
- djangoSkill
- event-driven-architectureSkill
- fastapiSkill
- graphqlSkill
- microservicesSkill
- nestjsSkill
- realtime-websocketsSkill
- spring-bootSkill
- competitive-analysisSkill
- lead-researchSkill
- product-analysisSkill
- resumeSkill
- user-researchSkill
- database-performanceSkill
- data-modelingSkill
- data-qualitySkill
- pandasSkill
- postgresSkill
- sparkSkill
- brand-guidelinesSkill
- generative-artSkill
- themingSkill
- ui-reviewSkill
- visual-designSkill
- architecture-decisionsSkill
- bug-fix-protocolSkill
- cli-developmentSkill
- code-reviewSkill
- debuggingSkill
- design-patternsSkill
- domain-modelingSkill
- git-workflowSkill
- legacy-modernizationSkill
- refactoringSkill
- repository-explorationSkill
- simple-designSkill
- system-designSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
2 rules. 0 are common, 2 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 9 edited versions we read keep these.
2
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 18
Before releasing a feature.
- Distinctive: line 58
Verify against the requirement, not against the implementation.
“Common” is measured: kept by at least 6 in 10 of the 9 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
QA Review
Purpose
Decide whether a change is ready to ship, on evidence. The output is a judgment with reasons — not "looks good to me", and not a list of every conceivable improvement.
When to Use
- Before releasing a feature.
- Reviewing a change for regression risk.
- Assessing whether the test coverage for a change is adequate.
- Sign-off on a release candidate.
Capabilities
- Acceptance-criteria verification against the actual behavior.
- Edge-case enumeration.
- Regression-risk assessment based on what the change touches.
- Test-coverage gap analysis.
- Release recommendation with explicit residual risk.
Inputs
- The change: the diff, the feature, and its acceptance criteria.
- The test evidence: what was tested, at what level, and what passed.
- The rollback capability.
Outputs
A release recommendation:
- Ship — Criteria met, risk understood and acceptable.
- Ship with mitigation — Ship behind a flag, or with a specific monitor in place.
- Hold — A specific, stated defect or gap must be closed first.
Each with the evidence behind it.
Workflow
- Verify the criteria, one by one — Against the running system, not against the pull request description. A criterion that cannot be verified was not a criterion.
- Enumerate the edges — For each input: empty, maximum, invalid, hostile. For each dependency: slow, down, wrong. Which of these are handled, and which were never considered?
- Assess the regression radius — What else touches the code that changed? A change to a shared utility has a far wider blast radius than a change to one route handler.
- Find the coverage gaps — Not "is coverage above 80%", but "is the new logic tested, and would the test fail if the logic were wrong?"
- Check the operational readiness — Can this be rolled back? Is it behind a flag? Will a failure be visible in monitoring, or will it be silent?
- Make the call, and state the residual risk — Shipping with known risk is legitimate. Shipping with unstated risk is not.
Best Practices
- Verify against the requirement, not against the implementation. The two agreeing proves only that the developer read their own code.
- A feature with no rollback path and no flag is a feature you are committing to permanently in the moment you deploy it. Say so.
- "The tests pass" is not release readiness. The tests pass on code that has never seen a real user.
- Identify the silent failure modes: what breaks in a way that no alert fires and no error is logged? Those are the ones that are discovered by a customer six weeks later.
- A hold must name a specific, closeable defect. "It feels risky" is not a hold; it is an anxiety, and it will be overruled.
- Record what was not tested. That is what makes the residual risk honest.
Examples
A release review that is a judgment, not a rubber stamp:
## Feature: partial refunds — release review
### Acceptance criteria
- [x] An admin can refund less than the order total. Verified: ord_01HX, $12 of $42.
- [x] The remaining refundable balance is displayed. Verified.
- [x] A refund cannot exceed the remaining balance. Verified: API returns 422.
- [ ] The customer receives an email confirmation. NOT IMPLEMENTED — the
email template exists but is not wired to the refund event. This is in
the acceptance criteria and is not in the change.
### Edge cases
- Zero-amount refund: rejected correctly.
- Refund exceeding balance: rejected by the API, but the UI shows a success
toast (defect #4412, fixed in this branch).
- Concurrent refunds: idempotency key prevents a double refund. Verified
with two simultaneous requests.
- Gateway timeout mid-refund: NOT TESTED. The gateway sandbox cannot be made to
time out. The code path exists and is unit-tested
with a mock, but has never run against the real
gateway.
### Regression radius
`OrderBalance` is shared with invoicing and with the accounting export. Both
have test coverage and both suites pass. The accounting export was manually
verified against a refunded order — the figures reconcile.
### Operational readiness
- Behind flag `refunds.partial` — default off. Rollback is a flag flip.
- A metric exists for refund failures; an alert does not.
### Recommendation: SHIP WITH MITIGATION
Ship behind the flag, enabled for internal accounts only, for one week.
Blocking for general availability:
1. The confirmation email (an acceptance criterion, not delivered).
2. An alert on the refund failure rate — without it, a gateway problem is
silent until a customer complains.
Residual risk accepted for the internal rollout: the gateway-timeout path is
untested against the real gateway. Impact is bounded to internal orders.Notes
- "Ship with mitigation" is the most useful and most under-used outcome. It unblocks the team while being honest about what is not finished.
- The most valuable line in most QA reviews is the "not tested" section. It converts an unknown risk into a known one, which is the entire job.
- A review that always says ship is not a review. A review that never says ship is an obstacle. The value is in the discrimination.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

