Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add denis-samatov/yandex-workspace-mcp --skill yandex-workspacegit clone --depth 1 https://github.com/denis-samatov/yandex-workspace-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/skills/denis-samatov/yandex-workspace-mcp/yandex-workspace)<a href="https://agentmods.dev/skills/denis-samatov/yandex-workspace-mcp/yandex-workspace"><img src="https://agentmods.dev/badge/skills/denis-samatov/yandex-workspace-mcp/yandex-workspace/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/skills/denis-samatov/yandex-workspace-mcp/yandex-workspace"><img src="https://agentmods.dev/badge/skills/denis-samatov/yandex-workspace-mcp/yandex-workspace.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 | $0.00015 | $0.00768 |
| Opus 5 | $0.00008 | $0.00384 |
| Sonnet 5 | $0.00003 | $0.00154 |
| Haiku 4.5 | $0.00002 | $0.00077 |
Grade A, and why
yandex-workspace 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 11d 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 — 28 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Yandex Workspace Skill
This skill instructs agents on how to best interact with the yandex-workspace-mcp server.
Best Practices
- Unified Search First: To find information across the workspace, always start with
searchto query both Yandex Disk and Yandex Wiki simultaneously. Use the returned locator/slug to fetch detailed content.- If a source reports
degraded=trueandsearch_mode="descendants", results match Wiki slugs from the bounded descendants fallback, not full page content. - Continue canonical searches only with the returned workspace cursor. Do not pass it to standalone Disk or Wiki tools.
- If a source reports
- Wiki as Documentation, Disk as Artifacts: Treat Yandex Wiki as the source of truth for structured documentation, project notes, and dynamic tables. Treat Yandex Disk as the source for raw files, datasets, and binary artifacts.
- Safe Wiki Updates: Use page write tools only after reading the target. For grids, pass the latest returned
revision; row, cell, and column mutations use optimistic concurrency. - Destructive Actions: Do not perform destructive actions (
wiki_delete_page,wiki_delete_grid, row/column deletion,disk_delete,disk_empty_trash, or overwrite moves) without explicit user consent. Treat a page recovery handle as short-lived and single-use. Emptying Trash is permanent and requires an explicit literal confirmation. - Conflict Handling: On a grid conflict, fetch the grid again, reconcile against its current revision, and ask before retrying a destructive change. Mutations are not automatically replayed.
- Allowed Roots Constraints: Be aware that your operations are confined to configured
ALLOWED_ROOTS. Do not attempt to traverse outside these roots, as the operations will be blocked. Results are post-filtered by the server even when the upstream API accepted a root filter. - Read Before Write: When asked to make mass changes, first fetch the list of objects (using
disk_list,wiki_get_descendants, or thewiki_get_treealias) and present a plan to the user before executing the write operations. - Attachments:
wiki_upload_attachmentcan only read a server-local file from an administrator allowlist and is available only in trusted stdio mode. Never assume a client-machine path is visible to the server. - Copy Operations: Page clone may wait for a bounded upstream operation. Grid copy can return
status="pending"; do not invent a completed grid ID when it does. - Disk Transfers: Prefer inline upload only for small generated text. Local-file/background tools exist only on a trusted stdio server and can see only administrator-allowlisted server paths. URL import accepts only administrator-allowlisted exact hosts. Never ask to weaken these gates.
- Local Upload Jobs: Poll only with the returned UUID using
disk_get_upload_statusor use the bounded list tool. Job records never contain the source path and disappear after their configured TTL or server restart. - Public and Trash Resources: A public key/URL must already be server-allowlisted and does not grant access to private roots. Trash results may omit entries whose original paths cannot be authorized; restore only to an allowed destination.
- Remote identity: Treat
workspace:read,workspace:write, andworkspace:deletedenials as server policy. Never ask for a Yandex token as an MCP bearer token or attempt to pass tokens, OAuth state, authorization codes, signed URLs, or public keys through tool text.
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.
- 11d ago First seen · 28 lines · 15 tokens per session scan A 126a1db4aefa
yandex-workspace is a skill published in the GitHub repository denis-samatov/yandex-workspace-mcp (0 stars, last pushed yesterday), licensed MIT. It adds 15 tokens to every session and 768 once invoked, about $0.0001 per session on Opus 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.
Other skills, from other repositories
auth-limits
Authenticate to Sugra API MCP and stay inside the daily request quota. Use when a tool returns 401, 429, missingapikey, or when setting up a client.
explore-catalog
Find and call the right Sugra API endpoint through the bundled catalog. Use when the agent does not already know the operationid, when a prompt recipe is the wrong shape, or before guessing parameter names.
hosted-vs-gateway
Choose hosted Sugra MCP versus the local gateway package, and which tools exist on each. Use when installing sugra-api-mcp, pointing a client at app.sugra.ai/mcp, or a hosted-only tool name fails on stdio.
cross-domain-briefing
Compose one briefing from two or three Sugra domains using only the eight gateway tools. Use when a question spans maritime, weather, macro, markets, or government and a single prompt recipe is not enough.
envelope-attribution
Parse Sugra API payloads, keep source attribution, and tell observation time from request time. Use when reading a callendpoint or fetchdata result, citing a figure, or shaping a large payload.
octave-compression
Workflow for transforming prose into semantic OCTAVE structures. Covers tier selection, transformation phases, loss accounting, and decision rules. REQUIRES octave-literacy.