Turn messy input into a sourced report. 15 default tools: workflow, bidirectional sync, canonical NodeBench research bridge, discover_tools, and load_toolset.
Scoring was not performed
24 of 44 tools have NO visible input schema in source code (findDocument, getDocumentContent, analyzeDocument, updateDocument, createDocument, and 19 others). Cannot verify parameter validation, types, or constraints.
Delegation tools (delegateToDataAccess, delegateToDocument, delegateToMedia, delegateToSEC, delegateToOpenBB) accept only a generic 'query' string parameter with no enum constraints, format hints, or guidance on expected query syntax. LLMs cannot distinguish between valid and invalid queries.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 29 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 38 | - | v1 |
invokeTool tool accepts untyped 'arguments' object with no schema validation. LLMs cannot know what fields are required, what types they accept, or what constraints apply. This is a critical composition anti-pattern.
Document management tools (findDocument, getDocumentContent, updateDocument, createDocument) lack documented output schemas. Agents cannot plan downstream operations or extract necessary IDs for tool chaining.
23 document and media tools have descriptions of 40 characters or fewer ('Search for documents by title and content', 'Read a document by ID and retrieve its full content').
No error handling documentation. No indication of retry guidance, error classification (retryable vs user-fixable), or recovery steps for any tools.
No documented output schemas for 24+ tools. Without output documentation, agents cannot parse responses, extract IDs for chaining, or validate data shape.
Destructive/write tools (createDocument, updateDocument, deleteDocumentEdit, createHashtagDossier) lack any mention of confirmation steps, dry-run modes, or rollback capabilities. No protection against accidental execution.
No pagination documented for tools that search or list (findDocument, searchHashtag, searchMedia, searchAvailableTools, listMediaFiles). Large result sets risk context window exhaustion.