Loading ToolBench
Single tool with complete input schema and basic description. Schema is well-formed with proper types, constraints (min/max), and required fields. Description is adequate (67 chars) but lacks context on when to use, prerequisites, or what happens to the file system. No output schema documented. Error handling returns text but lacks recovery guidance. Tool performs a write operation (WRITE risk) but description does not explicitly state this has side effects.
Capture a screenshot of any web page or local GUI
Output schema not documented. LLMs cannot plan downstream operations or extract structured data from screenshot results.
Description lacks state-change warning. Tool writes files to disk but description does not say 'Captures and saves a screenshot' or warn of side effects. Agents need to know this is irreversible.
Error handling returns plain text without recovery guidance. 'Screenshot error: <message>' gives LLM no actionable next step. Should suggest 'Check URL format' or 'Verify network access'.
outputPath parameter accepts arbitrary file paths. No validation against path traversal (e.g., '../../../etc/passwd'). LLM could be tricked into writing outside intended directory.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 51 | 2026-07-28+ | v2 |
No pagination or result limits documented. Tool returns single file path as text. If extended to batch screenshots, no guidance on handling large result sets.