Loading ToolBench
An MCP server for blockchain and DeFi portfolio management with support for multiple chains and wallet integration
Critical failure. Only 1 tool visible in source: 'openChrome' in backend/MCP/Server.py. The tool has an empty input schema {} with no parameters, a minimal description ('Open Chrome browser process'), and no documented output schema. The description is 28 characters, below the 34-character p10 baseline for A+ tools, and provides no context on WHAT happens, WHEN to use it, or what it returns. No error handling, no security guidance, and no confirmation mechanism for what appears to be a destructive/write operation (Risk: WRITE). The repository structure shows a large multi-chain/multi-contract blockchain application (Sui, Walrus, Hardhat, Flow, Filecoin) with extensive frontend (Next.js) and backend (Walrus/Express) dependencies, but the MCP server integration is severely underdeveloped. The tool definition appears to be a stub, likely auto-generated or incomplete. No evidence of schema validation, parameter constraints, or output structuring. This is a proof-of-concept at best, not production-ready.
Open Chrome browser process
Empty input schema. Tool 'openChrome' declares no parameters (input: {}) and no schema validation. Per HARD SCORING RULES, schema score MUST be 0.
Insufficient description (28 chars, below 34-char p10 baseline). 'Open Chrome browser process' lacks WHAT happens, WHEN to use it, prerequisites, and output format. Per HARD SCORING RULES, description score capped at 0-20 for <20-char descriptions; this is slightly higher but still well below actionable threshold.
No output schema documented. Tool is marked Risk: WRITE but returns no documented structure. LLM cannot plan downstream actions or extract returned process ID, port, or status.
Destructive operation (WRITE risk) lacks confirmation mechanism. Opening a Chrome process is a side effect, no dry-run, no confirmation step, no undo guidance.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 17 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 13 | - | v1 |
No error handling or recovery guidance. If Chrome is already running, unavailable, or launch fails, the tool offers no actionable error message or retry strategy.
Unclear naming intent. 'openChrome' is imperative and suggests a side effect, but lacks clarity on scope (local only? remote? headless?). Convention would be 'start_chrome_process', 'launch_browser', or 'open_browser_window' with explicit parameter control.
No security annotations or permission gates. A tool that spawns system processes should declare required permissions (e.g. 'execute:system', 'write:filesystem') and validate caller authority.