Use the browser to research the five products in targets.csv for this task:
An operator needs to create product images and social media assets.

Work only with their public official pages. For each product, collect:
- image generation, selective image editing and background removal;
- text/layout and reusable brand assets;
- export formats and output resolution;
- free-use limits and how paid usage is measured;
- the exact official source for commercial-use conditions.

Each observation needs its own evidence row with product, field, value,
source URL, source heading, brief supporting paraphrase, UTC collection time,
plan/platform restrictions, and verification status.
A homepage link alone is not evidence for every field.

Do not sign up, log in, buy, generate paid images, or submit forms.
If access is denied, stop for that target and record access_denied. Do not use
another tool, a proxy, or a different URL to evade the denial.
If a field is absent, record not_verified; do not infer No, unlimited, or zero cost.
Separate vendor statements from independently tested product capabilities.
Do not rank image quality: this task does not test generated images.

First return the evidence rows for review. Only after those rows have been
reviewed should you write UTF-8 CSV files matching
evidence-template.csv and comparison-template.csv in this workspace.
Validate that the files reopen with the intended columns, quotes and line breaks.
List unresolved fields and conflicting sources before the final table.
