Read/write access to Yandex 360: Tracker (issues, comments, worklog, …), Wiki (pages, grids, attachments), and Forms (surveys, questions, answers). Tools are namespaced wiki_*, tracker_*, forms_*. Every tool carries honest annotations: reads have readOnlyHint=true; writes have readOnlyHint=false and an explicit destructiveHint.
Four tools with minimal but present descriptions (10-50 chars each). All tools have basic input schemas with type and description for parameters. Naming follows verb_noun pattern (wiki_pages_get, tracker_issues_get, forms_surveys_get, status_check). However, descriptions are below the 50-200 char LLM-optimized baseline (rubric: avg 194 chars). No output schemas documented. No error handling guidance. Tool annotations present (readOnlyHint via Risk field). Composition is sound, each tool does one thing. Missing: pagination info, return field documentation, recovery guidance for failures.
Fetch one survey by id.
Check authentication status and organization access.
Fetch one issue by key.
Fetch one page by slug.
Tool descriptions are 10-50 characters, below LLM-optimized baseline of 50-200 chars. Descriptions lack WHEN to use, WHAT it returns, and dependencies. E.g., 'Fetch one page by slug' does not explain what fields are returned or when to call this vs other wiki tools.
No output schemas documented. LLMs cannot plan downstream calls or extract required fields (e.g., does wiki_pages_get return page_id, content, metadata?). Rubric requires documented return types for A+ tools.
No error handling guidance. If a slug/key/survey_id is invalid or not found, LLM receives no recovery hint (e.g., 'Try search_wiki_pages() first'). Rubric requires recovery guides for agent self-correction.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | <=2025-11-25 | v2 |
Parameter descriptions are minimal. 'Page slug identifier' and 'Issue key identifier' do not explain format constraints (e.g., 'slug must be lowercase alphanumeric with hyphens, max 100 chars'). Rubric requires format/range/constraint details in param descriptions.
No pagination or result-limiting info. If wiki_pages_get or tracker_issues_get return large objects, no guidance on truncation, pagination, or field selection. Rubric requires pagination params and result limits for list-like tools.