24 tools with consistent naming (geo_* prefix, verb-noun pattern) and complete input schemas. Descriptions present but generic (avg ~80 chars, below 194-char baseline). All parameters typed and described. Output schemas not documented. Error handling minimal. Missing tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels. No pagination guidance for list tools. Composition is sound, tools chain well via workspace/datastore/layer hierarchy.

• Document output schemas for all tools. For list tools, specify: returns array of {name, id, ...}. For create/update, specify: returns {id, name, status, created_at}. This enables agents to chain calls and extract IDs.
• Add tool annotations: mark geo_delete_* with destructiveHint=true, geo_list_* and geo_get_* with readOnlyHint=true, and idempotent tools with idempotentHint=true. Agents will then know which calls are safe to retry.
• Expand descriptions to 100-150 chars. Include WHEN to use each tool and dependencies. E.g., 'List all workspaces in the GeoServer instance. Call this first to discover workspace names before creating datastores or publishing layers.'
• Add pagination to list tools: accept optional limit (default 50, max 500) and offset/cursor parameters. Return total_count and next_cursor in responses. Document in descriptions: 'Returns up to 50 workspaces; use offset to paginate.'
• Add error recovery hints to descriptions. E.g., 'Delete a workspace. If deletion fails due to contained stores, set recurse=true or delete stores first. If workspace not found, call geo_list_workspaces to verify the name.'

Inferred effective spec: 2026-07-28+.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry