Cross-Service Patterns

Signal

Both services had `llms-full.txt` endpoints that returned 404 when tested. In Stripe's case, published `llms.txt` files reference the missing `llms-full.txt` URL as complete documentation (service-brief-backed). GitHub's `llms.txt` files (on both `github.com` and `docs.github.com`) do not reference `llms-full.txt` (live-verified 2026-04-24); the 404 was discovered by an agent probing the endpoint directly.

Why it matters

An agent following the llms.txt spec or probing the conventional `llms-full.txt` location may attempt to fetch it as a comprehensive context source. A 404 is a silent dead end — the agent doesn't know what it missed. Where the file is explicitly referenced from `llms.txt` (as in Stripe), this is worse than not referencing it at all, because it creates a false expectation.

What good looks like

Publish a working `llms-full.txt` at the expected location, or return a clear signal that no full-file variant is offered rather than a silent 404.

Signal

Developer docs use client-side JavaScript rendering. Agents fetching pages via curl/WebFetch get incomplete HTML, empty content, or 503 errors. Critical information (e.g., Stripe's declined card test tokens table, GitHub endpoint docs) is missing from the fetched response.

Why it matters

Agents cannot render JavaScript. If documentation relies on client-side rendering, agents lose access to content that exists but isn't in the initial HTML response. This is the single largest friction source in the Stripe audit.

What good looks like

Serve docs as static HTML or provide a documented machine-readable alternative (`.md` suffix, API endpoint, or complete llms-full.txt).

Mitigations observed

GitHub provides programmatic docs APIs (`/api/article`, `/api/article/body`) that agents used effectively during the audit. Stripe docs also accept a `.md` suffix returning clean markdown, but this was a post-audit discovery (service brief); the audited agents tried `.md` variants on affected pages without success. A live check on 2026-04-24 found `docs.stripe.com/testing.md` returning markdown and including declined-card test data, so the current page state appears better than the March audited runs.

Signal

At least two audited services publish `llms.txt` files that go beyond listing doc links. Stripe's docs-side file includes an explicit agent-instructions section. GitHub's file surfaces agent-relevant integration surfaces and programmatic docs APIs, though the stronger characterisation ("directs agents to use programmatic docs APIs rather than parsing HTML") is service-brief-backed rather than directly report-backed.

Why it matters

A plain link index helps agents find docs. Agent-specific instructions help agents make better choices about which docs to use and which API paths to prefer. This is the difference between discoverability and guided discovery.

What good looks like

Stripe's "Instructions for Large Language Model Agents" section; GitHub's "Building with GitHub" section directing agents to programmatic docs APIs.

Signal

Both services return structured error responses with machine-readable fields. GitHub error responses consistently included `documentation_url` linking to the relevant API docs. Stripe error responses included `type`, `message`, and `request_log_url`; the later rerun also showed `doc_url` on at least some error responses. The error response itself tells the agent where to learn more.

Why it matters

An agent encountering an error often gets either a documentation link or a request log, plus enough structured fields to recover without searching. This closes the loop between failure and recovery in a single response.

What good looks like

Stripe's error response format (`type`, `code`, `param`, `message`, `doc_url`); GitHub's format (`message`, `documentation_url`, `errors[]`).