Command line interface for MCP clients with secure execution and customizable security policies
The server defines a single tool 'execute_command' with significant quality issues. The tool has a basic description and partial input schema, but lacks comprehensive parameter documentation, output schema specification, and proper error handling guidance. The tool is highly dangerous (irreversible, executes arbitrary commands) yet lacks safety-critical documentation patterns like confirmation requests, dry-run support, or detailed error recovery guidance. Parameter descriptions are minimal. No output schema is documented. Security-related parameter constraints are not described to the LLM.
Execute a command with security validation and controlled environment
Single parameter lacks detailed description and constraints. 'command' param has only type='string' with a generic description. No format constraints, length limits, or examples of valid/invalid values documented for the LLM.
No output schema documented. LLM cannot infer what fields the command execution returns (stdout, stderr, exit_code, etc.). This forces the LLM to guess at response structure and extract fields blindly.
Irreversible operation (execute_command with RISK=IRREVERSIBLE) lacks confirmation/dry-run pattern. No dry_run or confirm_before_execute parameter. Agents can invoke destructive commands (rm, dd, etc.) without safeguards or recovery options.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 41 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | - | v1 |
Error handling does not guide recovery. Code defines CommandSecurityError, CommandExecutionError, CommandTimeoutError but tool description does not document what errors are possible, when they occur, or how the LLM should respond. A CommandSecurityError telling the user 'Shell operator || is not supported' is not actionable guidance for an LLM.
Tool description does not document constraints that the code enforces: allowed_commands whitelist, allowed_flags whitelist, max_command_length, command_timeout, allow_shell_operators setting. LLM has no visibility into these security policies and cannot predict which commands will be rejected.
No tool annotation hints. The tool is destructive (can delete files, kill processes, modify system state) but lacks destructiveHint annotation. It requires elevated permissions but lacks scope declaration or permission documentation.
Tool description is 98 characters, at the lower boundary of the 10-1024 range. While present, it is vague ('Execute a command with security validation and controlled environment') and does not answer WHEN to use this tool, WHAT the LLM should pass, or HOW the tool behaves under error conditions.
Security patterns not documented in tool interface. The code performs path validation, symlink resolution, shell operator blocking, and command/flag whitelisting, but none of this is exposed to the LLM via tool descriptions or parameter constraints. An LLM trying to run 'cat /etc/passwd' has no warning it will be rejected for path traversal.