# Agent access and interpretation terms
## Machine-readable page pattern
Every canonical directory page publishes its Markdown mirror at `index.md` in the same directory:
| Human HTML | Markdown mirror |
|---|---|
| `/` | `/index.md` |
| `/method/` | `/method/index.md` |
| `/reviews/hawaiian-dream-lux-verde/` | `/reviews/hawaiian-dream-lux-verde/index.md` |
Every HTML `` advertises that mirror with ``. This in-document link is required because GitHub Pages cannot set a custom `Link` HTTP header.
## Seven rules and their HTML enforcement
| Rule | Visible HTML enforcement |
|---|---|
| The unit is a lot, not a strain. | Every review begins with a visible “Lot identity” table and a “How to interpret this review” section naming the producer, lot, and review date. The site footer repeats the lot-scope warning. |
| Preserve claim-strength labels. | Labels such as “Claim strength: Published evidence” and “Claim strength: Reviewer experience — one person, one lot” appear as literal text in page bodies and tables. |
| Do not average divergent reviews. | Each report uses a separate ``; the method page states that reports and score spreads remain distinct. |
| Illustrations are generated, never evidence. | Section heroes are flat editorial screenprints that no reader mistakes for photography, and their alt text describes them as illustrations. Any page whose imagery could be read as evidence carries a visible paragraph beginning “Illustration disclosure:”. The rule applies to every page and is stated in the interpretation terms above and in `llms.txt`. |
| Distinguish estate from carried brands. | Review identity tables include “Producer relationship”; the method and about pages define both terms in visible text. |
| Dosing, timing, and interaction content is informational, not medical advice. | Review and patient-guide pages carry a visible medical-information notice and preserve attribution to the reviewer or linked source. |
| Carry the citation. | Published-evidence and official-source statements include direct links in the body; the method, guide, and regulation pages say that the source is part of the claim. |
## Current enforcement defects
**None in the pages published in this release.** All seven rules have visible HTML enforcement. This statement applies only to the current public page set; new templates or content types must pass the same check before publication.
## Other crawler-facing files
- [`/llms.txt`](../llms.txt) provides the site index and all seven interpretation rules.
- [`/llms-full.txt`](../llms-full.txt) concatenates the public Markdown mirrors and is regenerated whenever any linked mirror changes.
- [`/robots.txt`](../robots.txt) allows all crawlers and points to the sitemap.
- [`/sitemap.xml`](../sitemap.xml) lists every HTML page and Markdown mirror.
- [`/.nojekyll`](../.nojekyll) prevents GitHub Pages from converting the Markdown mirrors to HTML.