How real projects summarise themselves for language models.
jfarcandllms.txtjfarcand/mirroir-mcp/llms.txt
accept # re-record every baseline; refuses to run in CI ``` Three verdicts, three exit codes: **0 = PASS**, **1 = FAIL**, **65 = DRIFT** — every structural assertion held and a drift metric moved
roy-tongllms.txtroy-tong/AgentMeasure/llms.txt
claim. Verdicts are PASS / FAIL / UNPROVABLE. > Local-first. No network code in the tool. MIT. > > Last reviewed: 2026-09-19 ## What this standard is AgentMeasure is the **semantics layer
nirholasllms.txtnirholas/three.ws/public/llms-full.txt
# three.ws: Complete Page Index three.ws is the AI-agent layer for the open web
nirholasllms.txtnirholas/three.ws/public/llms.txt
# three.ws > three.ws is the AI-agent layer for the open web: build, embed, monetize
mihneaptullms.txtmihneaptu/opencode-fusion/site/llms.txt
them. Skills are discovered on demand; installing one needs no restart. 5. Cross-vendor review is emergent: pick main and sidekick from different model families. 6. Only the permission layer
garagonllms.txtgaragon/nanostack/llms.txt
work. Not a saved phase; the artifact appears on review. - /review: Two-pass code review (structural + adversarial). Detects scope drift against the plan and conflict-precedence against prior /security. - /security
KbWenllms.txtKbWen/agentic-os/llms.txt
agents. It gives agents — Claude Code, OpenAI Codex, Cursor, GitHub Copilot, Google Antigravity, and any Markdown-reading LLM agent — a repeatable plan → build → review → test → ship workflow with enforced quality
jenticllms.txtjentic/jentic-one/llms.txt
only the operations it has been approved for, and asking for more is a reviewable request rather than a silent widening. - Leave a trail. Every brokered call is recorded
nikolai-vysotskyillms.txtnikolai-vysotskyi/trace-mcp/docs/llms.txt
calls return vs making structural questions about code cheap to ask. - [trace-mcp vs code-review-graph](https://trace-mcp.com/vs/code-review-graph.html): Head-to-head against the incremental SQLite graph peer — blast
dylanroscoverllms.txtdylanroscover/Embody/web/llms-full.txt
bloom on the final output. ``` You're not getting a screenshot or a code snippet to paste. You're getting the **actual network**, in front of you, ready to play
KyaniteLabsllms.txtKyaniteLabs/kinocut/llms.txt
/KyaniteLabs/kinocut/blob/master/docs/360_ASSEMBLY.md - AI-video review (dev tip): https://github.com/KyaniteLabs/kinocut/blob/master/docs/AI_VIDEO_REVIEW_AND_SALVAGE.md - FAQ: https://github.com/KyaniteLabs/kinocut/blob/master/docs/faq.md - Agent skill: https://github.com/KyaniteLabs/kinocut/blob/master/skills/kinocut/SKILL.md - Changelog: https://github.com/KyaniteLabs/kinocut/blob/master/CHANGELOG.md ## Use cases - Claude Code / Cursor / Codex-style
tubecreatellms.txttubecreate/tubecli/docs/llms.txt
with no API key. Representative operations, all executed by agents rather than suggested as code: describe a site in plain language and the website agent runs `git clone` → `npm install
PsiACEllms.txtPsiACE/skills/skills/llms.txt
Reuse and composition](https://raw.githubusercontent.com/psiace/skills/main/skills/friendly-python/references/reuse-composition.md): Composition-oriented design. - [Review checklist](https://raw.githubusercontent.com/psiace/skills/main/skills/friendly-python/references/review-checklist.md): Code review focus areas. ## Piglet references (Markdown) - [Variables and naming](https://raw.githubusercontent.com/psiace/skills/main/skills/piglet/references/variables-and-naming.md): Naming, scope
alfadur7llms.txtalfadur7/llm-wiki-newsroom/docs/llms.txt
Claude Code that turns a folder of documents into a cross-linked, human-readable markdown wiki. The agent that writes a page is never the one that reviews
ihuzaifashoukatllms.txtihuzaifashoukat/x-use/llms.txt
Agent Skills support. - Draft mode is on by default. Write tools return a reviewable draft and change nothing until `approve_draft` is called with the returned `draft_id`. - Interactive
paolobietolinillms.txtpaolobietolini/gtm-mcp-server/llms.txt
template details including .tpl code | | `create_template` | Create a custom template from .tpl code | | `update_template` | Modify an existing template | | `delete_template` | Remove a template (requires `confirm: true`) | | `import_gallery
alexei-ledllms.txtalexei-led/spotinfo/llms.txt
docs/claude-desktop-setup.md) - [CLI and MCP surface review](docs/reviews/cli-and-mcp-surface-review.md) - [Surface validation](docs/reviews/surface-validation.md) - the built binary across the command x cloud x format matrix - [CLAUDE.md](CLAUDE.md) - AI coding guidelines for this project ## Commands
GuyMannDudellms.txtGuyMannDude/mnemo-cortex/llms.txt
facade, and ships drop-in integrations for the major MCP hosts (Claude Desktop, Claude Code, OpenClaw, LM Studio, AnythingLLM, Agent Zero, Hermes Agent, Open WebUI, llama.cpp, LobeChat, Jan). v3 (Bridge
Cassette-Editorllms.txtCassette-Editor/oh-my-cassette/llms.txt
trycassette.online) — beat-synced cuts, auto-matched music, subtitles, transitions, and picture-in-picture. Claude Code, Codex, Hermes Agent, and OpenCode launch the same local stdio MCP runtime. It handles media
JasonColapietrollms.txtJasonColapietro/suede-creator-skills/docs/llms.txt
positioning and campaign plans, or a map of operational bottlenecks. The pack also includes code review, app delivery, and creator-rights tools. Original work is MIT licensed; adapted components retain
A markdown file at a website's root that gives language models a short guide to the site and links to the pages worth reading.
The same idea with the content included, so an agent can read the documentation in one request.
Agents and tools that fetch documentation on someone's behalf. It's a proposal, not a standard, and support varies.
Start with a one-line summary, then sections of links with a sentence each. The examples here show what real projects do.
