---
title: Agent manifests (skill.md & AGENTS.md)
description: Auto-generated manifests that tell AI agents what your product is and how to read and edit your docs.
url: http://localhost:3040/guides/agent-manifests
---

# Agent manifests (skill.md & AGENTS.md)

Auto-generated manifests that tell AI agents what your product is and how to read and edit your docs.

Alongside `/llms.txt`, Thally generates two agent-facing manifests automatically —
no configuration required. Both describe your documentation *to agents*, and both
stay in sync with your content because they're generated from the same content
graph.

## `skill.md` — a capability manifest

Served at **`/skill.md`**, this is a Claude-skill-shaped description of what your
product is and how an agent can read your docs programmatically: the `llms.txt`
index, per-page Markdown (`.md`) and JSON endpoints, the search API, and your
OpenAPI spec. Point an agent at it and it knows how to work with your docs.

## `AGENTS.md` — guidance for editing the docs

Served at **`/AGENTS.md`**, this tells an agent *working on your repo* how the
project is laid out and how to add or edit pages safely (frontmatter rules,
`docs.json` registration, running `thally check`). It's also the config surface the
`thally agent` reads.

Drop a physical `AGENTS.md` at your project root to override the generated one
with hand-tuned guidance — style rules, files to never touch, review requirements.
The file wins over the generated default.

## Discovery

Both manifests are linked from `/llms.txt`, and Thally advertises the index on every
page with `Link: </llms.txt>; rel="llms-txt"` and `X-Llms-Txt` response headers,
plus a `/.well-known/llms.txt` mirror — so agents and crawlers find everything
without guessing.