Loading ToolBench
A repository containing Deepline plays, scoring systems, and example campaigns for GTM engineering workflows including route experiments, LinkedIn event organizer extraction, and target account warm-intro campaigns
This server exhibits severe definition quality gaps across all dimensions. Tool names lack consistent verb prefixes (e.g., 'route_experiment_judge' is unclear; 'linkedin_scraper_linkedin_search_events' is redundant). Descriptions are present but generic and lack actionable context for LLM selection. Input schemas are partially visible but inconsistent: some tools show full JSON Schema (route_experiment_judge, ai_inference), while others (linkedin tools) show minimal structure. No output schemas are documented. Error handling is absent from all tool definitions. The codebase shows TypeScript type definitions for internal data structures (RetrievedItem, RouteScore) but these are not exposed as tool output schemas. Tools 3-10 appear to be wrapper calls to external APIs (LinkedIn, Edges) with minimal validation or error guidance.
Execute AI inference with structured output for route experiment judging
Extract event details and organizer information from LinkedIn event URL via Edges API
Find LinkedIn profile URL by person full name and optional company name via Edges API
Search LinkedIn companies by name and return company profile information via Edges API
Search LinkedIn events by URL using Edges managed identity API
Extract company information from LinkedIn company search or profile URL
Inconsistent and unclear tool naming. 'route_experiment_judge' does not start with a clear action verb. 'linkedin_scraper_linkedin_search_events' is redundant (prefix repeated). LinkedIn tools mix snake_case and kebab-case. LLMs cannot reliably infer intent from these names.
No output schemas documented for any tool. Internal TypeScript types (RetrievedItem, RouteScore, RouteSelection) exist but are not exposed as tool return schemas. LLMs cannot plan downstream calls or extract required fields.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 44 | 2026-07-28+ | v2 |
Extract event details and organizer information from a LinkedIn event URL
Search LinkedIn events by URL and return event results
Find LinkedIn profile URL by person name and optional company name
Judge one fused route-experiment shortlist for task relevance.
LinkedIn tools (3-10) show minimal input schemas. Most expose only a single string parameter (url, linkedin_event_url, full_name) with no validation, constraints, or format guidance. No enum constraints, no length limits, no error recovery hints.
No error handling guidance in any tool definition. No indication of retryable vs. fatal errors, no recovery hints (e.g., 'If LinkedIn URL is invalid, try search_events first'), no categorization of failure modes.
Descriptions are generic and lack LLM-optimized context. Most are 35-50 characters, below the 50-200 char baseline for A+ tools. No WHEN-to-use guidance, no dependency hints, no distinction between similar tools (e.g., linkedin-search-events vs. linkedin_scraper_linkedin_search_events).