fact-sheet-search

Scope

This skill routes fact-sheet questions. Tool mechanics belong to tool descriptions; workflow details belong to the reference files. Treat every type, relation, field, facet, enum, tag, role, and example key here or in references as an example shape until it is confirmed in the current workspace. For each answer, bind the actual workspace keys from mcp__leanix__get_workspace_context, targeted tool results, or GraphQL introspection before building GraphQL. Semantic search is candidate discovery only; for broad concept routing rows, load the focused reference and verify candidates structurally before answering.

Retrieval tools for this workflow are mcp__leanix__get_workspace_context, mcp__leanix__filter_inventory, mcp__leanix__search_inventory, mcp__leanix__search_users, mcp__leanix__execute_code, and load_leanix_skill_reference. This skill already names the retrieval tools. If a retrieval tool is directly available, call it directly; if it is hidden, load its schema by exact name. Do not call mcp__leanix__search_mcp_tools in this workflow: it searches for tools, not inventory data. Do not use it to rediscover these known retrieval tools or to look up fact-sheet names, concepts, data objects, users, fields, lifecycle, ownership, or geography. Never satisfy a mandatory/safeguard/no-op step by calling mcp__leanix__search_mcp_tools with queries like "noop", "irrelevant", or "mandatory tool call"; skip that step and continue with the exact workflow tool. Avoid deprecated inventory shortcuts.

API Access

Before any LeanIX tool call: if only mcp__leanix__authenticate and mcp__leanix__complete_authentication are available, tell the user to run /mcp and authenticate the leanix server (browser opens automatically). Do NOT call authenticate yourself or suggest claude mcp add — the former returns a URL without triggering the browser flow (copy-paste UX), the latter would shadow the plugin's bundled server. If /mcp doesn't surface tools after auth, treat it as a plugin bug.

Tool availability & fallback

• mcp__leanix__execute_code fallback — when a workflow calls for mcp__leanix__execute_code (toolset code_execution) but it errors as unavailable, answer with one or more mcp__leanix__filter_inventory calls instead: push every predicate server-side as facet filters (FactSheetTypes, relation facets with target keys, lifecycle/tag/enum facets); for grouped counts use a filterOptions first:0 probe; for relations go target-first (resolve target IDs, then filter the source by relation facet keys:[targetIds]); chunk IDs across sequential calls when one filter.ids call is too large. If the answer genuinely needs in-memory joins, dedupe, or ranking across many pages that server-side facets cannot express, return the best server-side approximation with a short limitation note rather than presenting partial data as complete.
• Required tool not available — if a tool a question needs errors as unavailable and cannot be loaded by exact name, do not guess or fabricate. State plainly that the required tool(s) are not enabled, name them exactly, and recommend enabling the owning toolset (e.g. "The required tool mcp__leanix__execute_code (toolset code_execution) is not enabled in this session; ask your workspace admin to enable it, then retry."). For mcp__leanix__execute_code, apply the fallback above first and surface this only when the fallback cannot fully answer. If all inventory retrieval tools are missing, state the inventory workflow cannot run and name the inventory toolset plus tools.

Default flow