Lightweight agent-agnostic multi-agent orchestration platform
Four tools with basic schemas but critical gaps in descriptions and parameter documentation. All tools have input schemas with required fields, but descriptions are minimal (10-50 chars, well below the 194-char baseline). Parameters lack type constraints and detailed guidance. The shell__execute tool is particularly dangerous, no validation hints, no dry-run option, no permission gating. No output schemas documented. Error handling is absent from all tool definitions.
List files and directories in a workspace directory
Read contents of a file from the workspace
Write or create a file in the workspace
Execute shell commands in the workspace directory
shell__execute lacks any safety guardrails: no dry-run mode, no command validation hints, no permission checks, no output limits. Agents can execute arbitrary destructive commands (rm -rf /, kill processes, exfiltrate data) with no friction.
All tool descriptions are under 50 characters (baseline 194 chars). They lack WHEN to use, WHAT it returns, and any prerequisites. E.g., 'List files and directories in a workspace directory' does not explain when to call it vs other discovery tools, what fields are returned, or whether it follows symlinks.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields. E.g., does list_directory return file sizes, timestamps, permissions? Does read_file return encoding metadata?
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 35 | <=2025-11-25 | v2 |
filesystem__write_file has no overwrite/append mode parameter. Agents cannot safely update existing files, every write risks data loss if the file already exists. No idempotency guarantee.
No error handling guidance in any tool. If a file is not found, permission denied, or disk full, agents receive raw errors with no recovery hints. No categorization of retryable vs fatal errors.