Loading ToolBench
MCP server that fetches websites, converts them to PDF/Markdown documents or traverses links and returns URLs
Server defines 3 tools with complete input schemas and basic descriptions. However, descriptions are generic and lack LLM-optimized guidance on WHEN to use each tool or what distinguishes them. Tool names lack action verbs (convert, traverse, to-markdown are acceptable but not ideal verb_noun patterns). No output schemas documented. Parameters have types and descriptions but lack constraints (enums, ranges, patterns). No error handling guidance visible. Authentication via username/password parameters violates secret-injection pattern. Overall: functional but below production baseline.
Fetch a website and convert it to PDF document, with optional link traversal
Fetch a website and convert it to Markdown document, with optional link traversal
Traverse links on a website and return the URLs without creating PDFs
Credentials exposed as tool parameters (username, password). Violates secret-injection pattern, credentials in params leak into logs and agent traces.
No output schemas documented. LLMs cannot plan downstream operations or extract required fields (e.g., PDF buffer format, markdown content structure, URL list format).
Descriptions lack WHEN-to-use guidance. All three tools operate on URLs; descriptions do not clarify when to choose convert vs to-markdown vs traverse, forcing LLM to guess.
No input validation or error recovery guidance. If maxPages is negative, URL is malformed, or traversal times out, no guidance on what LLM should do next.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 43 | 2026-07-28+ | v2 |
Parameters lack constraints. maxPages accepts any number (no min/max); traverseLinks is boolean but no guidance on performance impact; no enum for authentication method.