Loading ToolBench
A Dart/Flutter-based AI assistant application with tool execution, MCP client integration, web search capabilities, and internal tool plugins for OCR, image generation, flowcharts, and more.
RemindAI has 16 tools with basic schemas and descriptions, but significant quality gaps. Tool names use inconsistent prefixes (toolshell_*, web_*) and lack clear verb-noun patterns. Most descriptions are under 100 chars and lack context on WHEN to use each tool or what happens on error. Parameters have type definitions but minimal guidance on constraints, ranges, or dependencies. No documented output schemas. Error handling is minimal, most tools lack recovery guidance. The parallel execution tool (toolshell_run_parallel) has complex logic but poor documentation of its safety constraints and conflict detection. Security-critical tools (write, delete, exec) lack explicit permission/audit documentation in their definitions.
Delete a file or directory
Execute a shell command
Read content from a file
Execute JavaScript code
Execute multiple tool calls in parallel (for independent operations)
Execute Python code
Create a checkpoint in the current worktree
Destructive tools (toolshell_delete, toolshell_exec, toolshell_run_python, toolshell_run_js) lack error recovery guidance. Descriptions do not explain what happens on failure, how to undo, or what the LLM should do next.
toolshell_run_parallel has complex conflict detection logic (resource conflicts, nested parallel prevention) but this is documented only in code comments, not in the tool description. LLMs cannot see code comments and will misuse the tool.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields (e.g., does toolshell_read return {content: string} or {status, content}?). This forces agents to guess and retry.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 48 | 2026-07-28+ | v2 |
Show diff in the current worktree
Finish a git worktree (merge or discard changes)
List all git worktrees
List all checkpoints in the current worktree
Revert worktree to a previous state (git reset --hard + clean)
Start a git worktree for isolated version workflow
Write content to a file
Fetch and extract readable content from a web URL
Search the web using DuckDuckGo (free, no API key required)
Worktree tools (toolshell_worktree_*) lack descriptions explaining when to use each one and how they relate. toolshell_worktree_checkpoint and toolshell_worktree_list_checkpoints have empty input schemas but no guidance on what a 'checkpoint' is or how to reference one.
Security-critical tools (toolshell_write, toolshell_delete, toolshell_exec) do not declare required permissions or audit requirements in their definitions. No indication that these require user confirmation or that they are logged.