Loading ToolBench

Reading the current corpus…

terminalOs MCP server · F · ToolBench

terminalOs

MCP server for remote server management via SSH, providing tools for host metrics, container management, key management, and command execution with access control policies

2026-07-28+13 tools

33/100

Definition Quality29

Protocol Readiness35

Supportability38

Spec posture

Inferred effective spec 2026-07-28+·current baseline 2026-07-28

Static source inference · medium confidence

Current-spec patterns detected

Summary

Critical gaps across all dimensions. Tool definitions exist but lack essential structure for LLM use. Descriptions are present but minimal (avg ~25 chars, well below 194-char baseline). No input schemas visible in source, only parameter names and types inferred from ToolRunner.swift logic. No output schemas documented. Parameters lack descriptions entirely. Error handling is generic. Tool names are reasonable (verb_noun pattern) but descriptions are too terse to guide LLM selection. This server is not production-ready for agent integration.

Tools (13)

add_hostwriteauthsource verified30/100

Завести сервер в списке

container_actionwriteauthsource verified27/100

start / stop / restart / rm

container_inspectread onlyauthsource verified30/100

Подробности контейнера

container_logsread onlyauthsource verified30/100

Последние строки логов

host_metricsread onlyauthsource verified30/100

Ядра, память, диск, сеть, аптайм

host_reportread onlyauthsource verified28/100

Что с хостом не так: пороги уже применены

list_authorized_keysread onlyauthsource verified30/100

Ключи на сервере

Issues (5)

CRITICALtool ↗

No input schemas visible in source code. Tool definitions in ToolCatalog.swift show only tool names and summaries; parameter schemas are inferred from ToolRunner.swift logic but not formally registered. LLMs cannot validate inputs or understand parameter constraints.

list_hostshost_metricshost_reportlist_containerscontainer_logscontainer_inspectlist_authorized_keysrun_commandcontainer_actionmanage_authorized_keyadd_hostupdate_hostremove_host

CRITICALtool-description ↗

Descriptions are too terse (avg ~25 chars, baseline 194 chars). Examples: 'Серверы и статус подключения' (28 chars), 'Ядра, память, диск, сеть, аптайм' (33 chars). Descriptions lack WHEN to use, WHAT it returns, and dependencies. LLMs cannot reliably select the right tool.

list_hostshost_metricshost_reportlist_containerscontainer_logscontainer_inspectlist_authorized_keysrun_command

Recommendations

Add formal JSON Schema input definitions to each tool. Include type, description, enum constraints, and min/max for all parameters. Example: host parameter should be {type: 'string', description: 'UUID of the target host (e.g., 550e8400-e29b-41d4-a716-446655440000)', pattern: '^[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}$'}.
Expand tool descriptions to 100-200 chars. Include WHAT (action), WHEN (use case), and WHAT IT RETURNS. Example for host_metrics: 'Retrieve CPU cores, memory usage, disk space, network stats, and uptime for a host. Call this to diagnose performance issues or monitor resource consumption. Returns structured metrics with thresholds applied.'
Add descriptions to every parameter. For 'action' in container_action, specify: 'Action to perform: start (begin stopped container), stop (gracefully halt), restart (stop then start), or rm (permanently delete). Destructive actions require confirmation.'
Document output schemas. Define what list_hosts returns (array of {id, name, user, address, port, status}), what host_metrics returns (object with {cores, memory_used, memory_total, disk_used, disk_total, network_in, network_out, uptime_seconds}), etc.
Enhance error messages with recovery guidance. Instead of 'не указан или не найден хост', return 'Host UUID not found. Call list_hosts to see available hosts, then retry with a valid UUID.' For permission denials, suggest 'This action requires confirmation. Ensure you have authority to modify this host.'

Score history

Overall score trend

First recorded score · v2 rubric

33/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	F	33	2026-07-28+	v2