Vibe GuideUse this guide to plan concrete tests and passing criteria, record actual results, then return to testing or repair.
Does this method fit your project, and how can you check?
- When would you use this method?
- When a feature can be used or a repair needs retesting.
- How should you choose for your situation?
- Write action → expected result from requirements, then record actual behavior. Separate automated checks, user trials and unexecuted checks; classify impact.
- What is easy to misunderstand?
- Appearance or an AI completion claim is not a test. Storage, login and multiple devices are not universal requirements; mark inapplicable cases.
- How can you check that this works for your project?
- Have someone reproduce the recorded steps. Preserve entry, version, steps, expected/actual results and errors; plan repairs before changing code.
TEST
Testing: check that it works
Testing compares expected and observed behavior through specific steps. Start with the main task, then check failures, other devices, and changes. Record AI checks separately from hands-on trials.
Move forward using observed results
1. Complete the normal flow
Use test data to follow the requirements from start to result; check saving and reopening only when persistence is needed. For example, a reservation flow selects a date, enters details, submits, and shows an ID.
How can you tell this item is complete?The task can be completed without the author explaining it; success matches a real record.
2. Try failure cases
Try empty required fields, an invalid date, repeated clicks, offline use, and declined permissions, one at a time.
How can you tell this item is complete?Explain the problem and recovery route without unexpected data loss or duplicates; explain inapplicable scenarios.
3. Change device and identity
Check text, buttons, and persistence on target phones and computers; use two identities for private-data access.
How can you tell this item is complete?Record actual devices and versions; browser resizing does not prove every real device works.
4. Retest after changes
Ask AI to list affected flows, retest the original issue, nearby behavior, and one complete task.
How can you tell this item is complete?Record this version and observed results; an earlier passing version is not evidence for this one.
Checks are personal marks on this page, not automatic tests or saved records. Reloading clears them.
Write check results
Edit the template below and confirm to prepare a copyable draft. Use it as directed by this step; sending it to AI is not required.
Confirmed · personalized content ready to copy
Version and device: [fill in actual record or unverified]
Task goal: [fill in actual record or unverified]
Test data: [fill in actual record or unverified]
Steps: [fill in actual record or unverified]
Expected result: [fill in actual record or unverified]
Actual result: [fill in actual record or unverified]
Evidence location: [fill in actual record or unverified]
Reproducible: [fill in actual record or unverified]
Severity and affected users: [fill in actual record or unverified]
AI check / hands-on trial: [fill in actual record or unverified]
Unverified items and next step: [fill in actual record or unverified]Confirm and copy into project notes; the draft is stored in this browser. Mark unrun checks unverified, not passed.
Open troubleshooting
Choose the next step from the result
Data disappears after reload
Check for the test record on another device using the same identity.
- Found → inspect retrieval, identity, and cache on this page, then reload and retest.
- Not found → inspect write success and storage location; preserve IDs and errors.
Exit: rerun the original steps. On success, record evidence and continue. Otherwise retain results for further investigation and keep the task incomplete.
Clicking has no effect
Does clicking show a loading or error state?
- Yes → record the full response and ask AI to locate the request or validation failure; retest.
- No → check disabled state and action wiring; verify the real result once connected.
Exit: rerun the original steps. On success, record evidence and continue. Otherwise retain results for further investigation and keep the task incomplete.
Cramped mobile layout
Does the target width produce page-level horizontal scrolling?
- Yes → provide width and screenshot, locate the oversized element, then retest long text.
- No → try tapping and typing, record awkward spacing or obstruction, then adjust.
Exit: rerun the original steps. On success, record evidence and continue. Otherwise retain results for further investigation and keep the task incomplete.
Access denied after sign-in
Should the current identity have this access under the agreed rules?
- Yes → inspect session and server rules; do not disable access checks.
- No → rejection is expected; provide a return route and owner contact.
Exit: rerun the original steps. On success, record evidence and continue. Otherwise retain results for further investigation and keep the task incomplete.
Another person’s data is visible
Can an unrelated test identity still retrieve that record?
- Yes → pause the affected feature, fix server rules, and retest both identities.
- No → check whether the original identity is an admin or shared member; explain the actual rule.
Exit: rerun the original steps. On success, record evidence and continue. Otherwise retain results for further investigation and keep the task incomplete.
Input is lost offline
Does the agreed scope promise offline persistence?
- Yes → reproduce reopening with test data; fix local persistence and sync conflicts, then retest.
- No → explain offline limits and preserve current input; return to scope confirmation before adding offline features.
Exit: rerun the original steps. On success, record evidence and continue. Otherwise retain results for further investigation and keep the task incomplete.
Repeated submission creates duplicates
Do repeated clicks on one action create different record IDs?
- Yes → inspect server deduplication and submission state; disabling a button alone needs repeated-request verification.
- No → inspect duplicate display and compare the actual stored count.
Exit: rerun the original steps. On success, record evidence and continue. Otherwise retain results for further investigation and keep the task incomplete.
Loading never finishes
Does the agreed waiting period end with an error and retry route?
- Yes → preserve error and time; inspect service status or request results.
- No → add timeout, cancellation, or retry feedback and simulate failure to retest.
Exit: rerun the original steps. On success, record evidence and continue. Otherwise retain results for further investigation and keep the task incomplete.
The release cannot be opened
Is even the home page unreachable?
- Yes → inspect address, running service, and domain status before application code.
- No → inspect the failing route, assets, and requests; retest from the live entry.
Exit: rerun the original steps. On success, record evidence and continue. Otherwise retain results for further investigation and keep the task incomplete.
Old data fails after an update
Can the preserved old version read a test copy?
- Yes → inspect format changes and validate migration or compatibility before updating.
- No → stop overwriting and inspect copy and backup integrity; insufficient evidence remains unresolved.
Exit: rerun the original steps. On success, record evidence and continue. Otherwise retain results for further investigation and keep the task incomplete.
AI forgets requirements
Ask AI to restate the latest scope. Does it match the approved material?
- Yes → identify the specific deviation, request a focused correction, and recheck.
- No → provide the latest material and exclusions, then confirm understanding before continuing.
Exit: rerun the original steps. On success, record evidence and continue. Otherwise retain results for further investigation and keep the task incomplete.
Repeated fixes fail
Are reproducible steps and observed evidence available?
- Yes → summarize attempts and results, compare hypotheses, and test one cause at a time.
- No → collect device, version, steps, and exact errors; pause speculative broad changes.
Exit: rerun the original steps. On success, record evidence and continue. Otherwise retain results for further investigation and keep the task incomplete.
Fill in the current situation to prepare a prompt for AI
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
Which version will you check, and where will you open it: [fill in]
Confirmed requirements and passing criteria: [fill in or paste]
What data and devices will you use for testing: [fill in]
Existing AI check record: [fill in or not run]
Prepare a hands-on test checklist only, without changing code or marking unrun tests passed. Include action, expected result and recording location.
Explain your conclusions using these points: Core flow, necessary failures and regression checks with outcomes left for actual testing. Produce reproducible issue records on failure, without fixing yet.