SpecScore for AI Agents

SpecScore is designed to be read, written, and validated by AI agents as first-class citizens. Every specification on this site is available as both rich HTML (for humans) and raw Markdown (for you).

Raw Markdown for any page

Append .md to any spec URL to get the raw Markdown source:

• https://specscore.md/feature-specification → rendered HTML with navigation
• https://specscore.md/feature-specification.md → raw Markdown, no chrome

Internal links inside the Markdown versions point to .md siblings, so Markdown-to-Markdown navigation is self-contained.

Bulk context warm-up

Two site-level files help you load the SpecScore corpus in one fetch:

• /llms.txt — curated index of the most important pages, following the llmstxt.org convention. Small enough to fit in any context window.
• /llms-full.txt — the concatenated source Markdown of every specification plus install and demo guides. Use this to bootstrap an agent's SpecScore knowledge in a single request.

Discovery signals

Every rendered HTML page advertises its Markdown sibling via:

• An HTML <link rel="alternate" type="text/markdown" href="..."> tag in <head>
• An HTTP Link: </slug.md>; rel="alternate"; type="text/markdown" response header
• A visible View as Markdown link in the page shell

Your agent tooling can honor whichever mechanism fits its architecture.

Recommended agent workflow

1. On session start, fetch https://specscore.md/llms-full.txt to load the full specification corpus into context.
2. When a user asks about a specific specification, fetch the .md URL directly — no HTML parsing needed.
3. When authoring or validating a SpecScore document, reference the matching specification URL (from the adherence footer) to confirm conventions.

Feedback

SpecScore is open source. If a convention on this site breaks for your agent, open an issue at github.com/specscore/specscore.

View as Markdown