Loading ToolBench
Code intelligence MCP server for AI coding agents — semantic code search, symbol navigation, and cross-language bridge detection
ministr provides 4 tools with complete input schemas and descriptions. Naming follows verb_noun convention (ministr_run, ministr_run_logs, ministr_run_status, ministr_run_kill). Descriptions are present and substantive (80-150 chars), explaining WHAT each tool does and WHEN to use it. However, parameter descriptions lack actionable constraints (e.g., timeout_secs capped at 3600 is mentioned in description but no min/max schema constraints visible; background flag lacks guidance on when to use it). Output schemas are not documented, LLMs cannot predict response structure for downstream chaining. Error handling is minimal; no recovery guidance or error categorization visible. Risk annotations (IRREVERSIBLE, READ_ONLY, REVERSIBLE) are present but not formalized as tool annotations in the schema. Composition is sound: each tool has one responsibility, and run_id chaining is clear.
Execute a command in a configured workspace and capture output with structured diagnostics
Cancel a running command execution
Retrieve or search captured output from a run, with delta-mode pagination that never re-sends already-delivered spans
Get the current status snapshot of a run without retrieving logs
Output schemas not documented. LLMs cannot predict response structure (fields, types, pagination) for ministr_run_logs, ministr_run_status, or ministr_run. This breaks downstream tool chaining and forces agents to guess field names.
Parameter constraints not formalized in schema. timeout_secs capped at 3600 is mentioned in description text but no JSON Schema maxValue constraint visible. max_bytes default 16KB mentioned but no schema bounds. LLMs cannot validate inputs before calling.
Risk annotations (IRREVERSIBLE, READ_ONLY, REVERSIBLE) are informal comments, not formalized as tool annotations (destructiveHint, readOnlyHint, idempotentHint). MCP spec supports these; using them would enable client-side safety guardrails.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 62 | 2025-06-18+ | v2 |
No error handling guidance. If ministr_run times out or fails, no recovery hint provided. If run_id is invalid, no suggestion to call ministr_run_status first. Agents cannot self-correct.
Parameter descriptions lack actionable detail. 'background' flag has no guidance on when to use it (e.g., 'Set to true for long-running commands to avoid blocking the agent'). 'query' in ministr_run_logs lacks format hints (regex? substring? case-sensitive?).