# Verification brief: one observable feature change

Adapt this template to an authorized development environment. This file does not
run a model, connect an account, or grant permission for any action.

## Scope and owner

- Change/job:
- Repository and source revision:
- Environment/fixture version:
- Reviewer/decision owner:
- Allowed actions:
- Excluded actions and data:

## Starting state

- Prerequisites:
- Fictional inputs/fixture:
- Reset or recovery procedure:
- Credentials: use the existing approved mechanism; never paste secrets here.

## Check

1. Open:
2. Perform this exact action:
3. Expected visible result:
4. Expected application/state result, if relevant:
5. Independently check the boundary or failure case:

## Evidence (complete only after running)

- Execution time and environment:
- Source revision actually tested:
- Exact non-sensitive input:
- Observed visible result:
- Observed state/result:
- Evidence artifact or reproducible command:
- Result: NOT RUN / PASS / FAIL / BLOCKED
- Limitation or missing evidence:

## Separate checks

- Unit/logic check: NOT RUN
- Visible/browser acceptance check: NOT RUN
- CI for the exact source revision: NOT RUN
- Review: NOT RUN
- Human decision: PENDING

A pass at one level does not fill in another. A prepared brief is not a test run,
a CI pass is not user acceptance, and ready for human review is not permission to
merge or deploy. Keep customer data, tokens, and private documents out of shared
evidence.

## Synthetic example

Fixture: https://rida.me/verification-kit/fixture.html?case=regression
Input: 2 seats, 25 credits each.
Expected visible result: 50 credits.
Deliberate regression: the calculation returns 50 but the interface displays 25.
Repeat with ?case=corrected to compare the same acceptance check.

This is a deterministic exercise, not a customer result or an agent-run claim.