Borrowing it
Nothing to install: this file belongs to eschatus/beyond-mcp. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/eschatus/beyond-mcp/main/CLAUDE.mdgit clone --depth 1 https://github.com/eschatus/beyond-mcpWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/eschatus/beyond-mcp/claude-md)<a href="https://agentmods.dev/instructions/eschatus/beyond-mcp/claude-md"><img src="https://agentmods.dev/badge/instructions/eschatus/beyond-mcp/claude-md/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/instructions/eschatus/beyond-mcp/claude-md"><img src="https://agentmods.dev/badge/instructions/eschatus/beyond-mcp/claude-md.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.01915 | $0.01915 |
| Opus 5.5 | $0.00766 | $0.00766 |
| Sonnet 5.5 | $0.00383 | $0.00383 |
| Haiku 4.5 | $0.00192 | $0.00192 |
Grade A, and why
beyond-mcp CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 21d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 115 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
What this is
Read-only D&D Beyond lookup MCP server, extracted from roll20-dm-mcp (its issue #171: reduce that repo to Roll20 primitives, put DDB behind its own seam). Consumers: the dm-whisper gem (live sessions) and Claude Code via the module-ingest skill (prep time). It serves character /monster/campaign lookups and captures live game-log dice rolls into a drainable buffer.
Invariants — do not violate
- Reads are always on; writes are opt-in and are primitives. This server can effect
changes on D&D Beyond (Adam's call, 2026-08-25, reversing the earlier read-only stance).
Write tools register ONLY when
DDB_WRITES=enabled, so prep-time / default installs stay read-only. The write layer (src/bridge/ddb-write.ts,src/tools/ddbWrite.ts) is a primitive surface: it effects exactly the change asked on whatever character id it's given, with NO policy about who owns the sheet or how the server is used upstream — the caller owns intent (this is a general MCP server, ignorant of its consumers). DM-authority model: DDB PC HP is also driven by Beyond20 and the player's own edits, so writes race those (last-writer-wins) — they are deliberate acts, never an automation loop on PC HP. Every write route's body shape is verified live against character-service before its tool ships (contract-probe with an empty body; the 400 names required fields) — never add a wrapper for an unverified route. - Homebrew creation rides the same gate (
src/bridge/ddb-homebrew.ts,src/tools/ddbHomebrew.ts). There is no homebrew API: it drives www's server-rendered builder forms over plain HTTP (rtSiteSessioncookie jar,html-form.tsparse/fill/encode) — still no browser. Rules that are easy to break: address controls by stable element id (names are hashed per render), match selects by label only (values collide with labels — CR "5" is value 9), treat a redirect to/erroras failure, and verify every write by read-back (edit form + monster-service + baked label tables). A new homebrew type ships only after a live round-trip on a throwaway private creation. Builder reads lag writes: verify throughsettle(), confirm deletes by a repeat delete answering 404, resolve edit paths withresolveEditPath. Deletes are SOFT (creations stay listed as Deleted in My Creations), so name test creationszz beyond-mcp …. Option-id escape for ambiguous labels is{ value }everywhere (never a bare number). Protocol:docs/ddb-homebrew-protocol.md; per-type notes:docs/ddb-homebrew-types.md. - Never open a browser. The CobaltSession cookie is consumed from
DDB_COBALTor<data dir>/ddb-cobalt.json; the interactive harvest UX lives in the gem. When the cookie is absent/expired, fail loudly with re-harvest instructions (CobaltUnavailableErrorin ddb-rt.ts) — never add a silent fallback or a Playwright dependency. - No VTT knowledge. Nothing Roll20- (or Foundry-) specific belongs here. The roll pump
buffers raw game-log messages; rendering into a VTT (e.g.
post_roll_as_characterin roll20-dm-mcp) is the consumer's job. Keep helpers VTT-neutral (summarizeRoll). - Tool shapes are a contract.
ddb_get_character/ddb_get_monster/ campaign listing default outputs match roll20-dm-mcp's originals so callers migrate seamlessly. Extend additively (like thefullflags), never reshape defaults. - Rolls captured by the pump carry DDB's actual dice values — consumers must mirror them, never re-roll. Keep tool descriptions saying so.
- Version is a consumer contract.
ddb_pingand MCPserverInfoboth announcesrc/version.ts'sVERSION; consumers pin a ref and floor the version against it (the gem'sEXPECTED_DDB_MCP_VERSION). So: bumpVERSIONandpackage.jsontogether on any consumer-visible change, and cut avX.Y.Zgit tag on the release commit (roll20-dm-mcp convention) so consumers pin#vX.Y.Z, not a bare SHA. Adding a tool is a minor bump.git pushdoes NOT push tags —git push origin vX.Y.Zexplicitly.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 21d ago Changed · +22 lines · +439 tokens per session b7456cd3e5a5
- 2mo ago First seen · 93 lines · 1,476 tokens per session scan A 45e194f7e04a
beyond-mcp CLAUDE.md is an instructions file published in the GitHub repository eschatus/beyond-mcp (0 stars, last pushed 19d ago), licensed MIT. It adds 1,915 tokens to every session, about $0.0077 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
beyond-mcp CLAUDE.md compared with instructions from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| beyond-mcp CLAUDE.mdeschatus/beyond-mcp | Instructions file | 0 | 19d ago | MIT | A | 1,915 | Claude Code |
| vscode buildNext.instructions.mdmicrosoft/vscode | Instructions file | 195k | yesterday | MIT | A | 9,719 | GitHub Copilot |
| next.js AGENTS.mdvercel/next.js | Instructions file | 143k | yesterday | MIT | A | 7,461 | Codex, OpenCode |
| langchain AGENTS.mdlangchain-ai/langchain | Instructions file | 149k | yesterday | MIT | A | 4,680 | Codex, OpenCode |
| vscode oss-third-party-notices.instructions.mdmicrosoft/vscode | Instructions file | 195k | yesterday | MIT | A | 5,516 | GitHub Copilot |
| vscode copilot-instructions.mdmicrosoft/vscode | Instructions file | 195k | yesterday | MIT | A | 3,093 | GitHub Copilot |
| deepseek-harness AGENTS.mddeepseek-ai/deepseek-harness | Instructions file | 274k | 2d ago | MIT | A | 4,313 | Codex, OpenCode |