Loading ToolBench
Lightweight terminal API for AI agents - provides file operations, command execution, and interactive terminal sessions
Open Terminal Lite provides 11 tools with complete input schemas and descriptions. Naming follows verb_noun convention (get_cwd, list_files, execute_command). Descriptions are present but brief (avg ~80 chars, below the 194-char baseline). Parameters have types and descriptions, but lack constraints (enums, ranges, patterns). Output schemas are not documented. Error handling exists but lacks recovery guidance. No tool annotations (readOnlyHint, destructiveHint) despite clear risk levels. Security concerns: execute_command accepts arbitrary shell commands with minimal validation; delete_file and move_file are destructive but lack confirmation patterns.
Delete a file or directory.
Execute a shell command and return output.
Get the current working directory.
List directory contents with detailed metadata (size, permissions, modification time). Similar to ls -la.
Create a directory (and parent directories if needed).
Move or rename a file or directory.
Read the contents of a file. Supports line ranges and PDF extraction.
execute_command lacks input validation and safety constraints. No enum for allowed commands, no blocklist for dangerous operations (rm -rf, dd, etc.), no timeout default. Description does not warn of irreversible consequences.
Destructive tools (delete_file, move_file) lack confirmation/dry-run patterns. No tool annotation hints (destructiveHint) to signal risk to LLM. Descriptions do not mention irreversibility.
Output schemas not documented. LLMs cannot infer what fields list_files, read_file, or search_files return. Prevents downstream tool chaining and forces agents to guess field names.
Parameter descriptions lack constraints. 'timeout' in execute_command has no min/max (could be 0 or 999999). 'path' parameters lack format guidance (absolute vs relative, symlink handling). No enums for case_sensitive or other boolean flags.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 62 | 2026-07-28+ | v2 |
Find and replace exact strings in a file.
Search for files matching a pattern or containing specific text.
Change the current working directory.
Write text content to a file. Creates parent directories automatically.
Error handling returns generic messages ('Path escapes sandbox', 'File not found') without recovery guidance. No suggestion of alternative tools or next steps for LLM.