Borrowing it
Nothing to install: this file belongs to A1-x-Tech/mcp-yandex-audience. 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/A1-x-Tech/mcp-yandex-audience/main/CLAUDE.mdgit clone --depth 1 https://github.com/A1-x-Tech/mcp-yandex-audienceWrote 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/a1-x-tech/mcp-yandex-audience/claude-md)<a href="https://agentmods.dev/instructions/a1-x-tech/mcp-yandex-audience/claude-md"><img src="https://agentmods.dev/badge/instructions/a1-x-tech/mcp-yandex-audience/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/a1-x-tech/mcp-yandex-audience/claude-md"><img src="https://agentmods.dev/badge/instructions/a1-x-tech/mcp-yandex-audience/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 | $0.02759 | $0.02759 |
| Opus 5.5 | $0.01104 | $0.01104 |
| Sonnet 5.5 | $0.00552 | $0.00552 |
| Haiku 4.5 | $0.00276 | $0.00276 |
Grade A, and why
mcp-yandex-audience 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 2mo 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 — 152 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md — mcp-yandex-audience
MCP server for the Yandex Audience API (TypeScript, stdio). Write-capable:
tools wrap segment management (file/CRM uploads with two-phase confirm,
lookalike, pixel-based), pixels and access grants; raw_request is the escape
hatch. The server talks to the Yandex Audience Management API
(api-audience.yandex.ru, /v1/management/*); auth is a Yandex OAuth token
sent as Authorization: OAuth <token> on every request — either from
YANDEX_AUDIENCE_TOKEN or minted by the in-chat login (start_login /
finish_login).
Commands
npm run dev # run from source (tsx watch)
npm test # unit tests + dist smoke (incl. real stdio MCP handshake), no network
npm run typecheck # types for src + tests
npm run build # emit dist/
npm run smoke # live READ-ONLY call (needs YANDEX_AUDIENCE_TOKEN)
Architecture
src/config.ts— env → config. A missingYANDEX_AUDIENCE_TOKENis NOT an error: the field staysundefined, the server starts and the token is resolved per request by theTokenStore.ConfigError(with areasoncode) is reserved for malformed values, caught byloadConfigOrDegradedinindex.ts(no such checks exist today). OptionalYANDEX_AUDIENCE_API_HOST(defaulthttps://api-audience.yandex.ru),YANDEX_AUDIENCE_TIMEOUT_MS,YANDEX_AUDIENCE_MAX_RETRIES.src/oauth.ts— the OAuth flow: PKCE pair (S256), authorize URL with thehttps://oauth.yandex.ru/verification_coderedirect, code exchange and refresh. Noclient_secret— this is a public client, and a secret inside an npm package would protect nothing. Scope is pinned toaudience:read audience:write. The pending verifier lives in one module-level slot (one stdio server = one user); a secondstart_loginreplaces it. Own app viaYANDEX_AUDIENCE_OAUTH_CLIENT_ID.src/credentials.ts—~/.config/mcp-yandex-audience/credentials.json, mode0600. An unparsable file reads as "not connected", never as an empty token.src/auth.ts—TokenStore: resolves the token per request (env wins over stored), refreshes on expiry, and raisesAuthRequiredErrorwhose message is the product — it is the only text the user ever sees about a missing token, and it names both fixes (start_loginin the chat, orYANDEX_AUDIENCE_TOKEN+ restart). It replaced the oldCredentialsError.src/client.ts— one typed method per endpoint (listSegments,confirmSegment,createLookalikeSegment,createPixel,addSegmentGrant, …): OAuth header, path/query/body assembly ({segment: …}/{pixel: …}/{grant: …}envelopes,compact()drops undefined),upload()for multipart file uploads (no explicit Content-Type — fetch sets the boundary). The token comes from theTokenStoreon EVERY attempt (never cached on the instance); a missing token raisesAuthRequiredErrorwhile the auth header is built — rethrown before the retry/backoff branch and before fetch. A 401 (or 403 naminginvalid_token) on a stored token triggers one silent refresh + replay.request()resolves the path against the base and rejects any path that escapes to a foreign origin (SSRF guard), enforces an AbortController timeout that also covers reading the body, and throwsAudienceError(status, body)parsing the API's{errors: [{error_type, message, location}], code, message}format. Retries: 429 for any method; 5xx/network only for GET — repeating a committed write would duplicate it.src/tools/auth.ts—auth_status,start_login,finish_login,logout: the in-chat login.finish_loginstores the token and proves it with a livelistSegmentsread;logoutdeletes only the stored file, never the env token.src/tools/segments.ts—list_segments,upload_segment_file,upload_segment_csv_file,confirm_segment,rename_segment,delete_segment,create_lookalike_segment,create_pixel_segment. Upload tools acceptfile_pathXORcontent(checked in the handler — cross-field rules can't live in a plain-object inputSchema).src/tools/pixels.ts—list_pixels,create_pixel,update_pixel,delete_pixel.src/tools/grants.ts—list_segment_grants,add_segment_grant,delete_segment_grant.src/tools/raw.ts—raw_request(GET/POST/PUT/DELETE, defaults to GET).src/tools/util.ts—ok/fail, the annotation constants (READ_ONLY/WRITE/WRITE_IDEMPOTENT/DESTRUCTIVE/RAW) and shared zod field factories.src/index.ts— wires everyregister*into the McpServer.loadConfigOrDegradedstarts the server even on a config problem;connected = tokens.hasToken()is resolved once, only to pick the instructions text — without a token they open with the unconfigured prefix (connect viastart_loginright in the chat, or setYANDEX_AUDIENCE_TOKENand restart).src/telemetry.ts— anonymous usage pings (ids/names/versions only, never data or arguments; fire-and-forget, must never block or throw; opt-outASKADS_TELEMETRY=0).server_startmeans "a usable install started"; an install without a token sendsunconfigured_startinstead. Thereasonis a closed vocabulary (missing_token) — never a variable's name or value.startup_failedremains for a config unusable at load time (malformed values; no such checks exist today), also fire-and-forget.
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.
- 2mo ago First seen · 152 lines · 2,759 tokens per session scan A d89d92e6aa1a
mcp-yandex-audience CLAUDE.md is an instructions file published in the GitHub repository A1-x-Tech/mcp-yandex-audience (0 stars, last pushed 6d ago), licensed MIT. It adds 2,759 tokens to every session, about $0.0110 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.
mcp-yandex-audience CLAUDE.md compared with instructions from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| mcp-yandex-audience CLAUDE.mdA1-x-Tech/mcp-yandex-audience | Instructions file | 0 | 6d ago | MIT | A | 2,759 | 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 |