Loading ToolBench

Reading the current corpus…

genpark-agent-sliding-context-window-skill MCP server · F · ToolBench

← Browse

genpark-agent-sliding-context-window-skill

github.com/alphaparkinc/genpark-agent-sliding-context-window-skill

MCP server implementing a sliding context window skill for managing and trimming message histories

2026-07-28+STDIO1 tools

12/100

Definition Quality7

Protocol Readiness5

Supportability25

Spec posture

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

Static source inference · medium confidence

Current-spec patterns detected

Summary

The 'execute_skill_action' tool has a vague, non-actionable description ('Execute a skill action (declared in mcp_server.py supported_tools)'). No input schema is visible in the source code, the tool definition is inferred rather than explicitly registered with JSON Schema. The tool name 'execute_skill_action' uses generic verbs ('execute', 'action') that lack specificity. No parameter descriptions, type definitions, or output schema are documented. The server prints a JSON handshake but does not implement the MCP protocol, it outputs a single JSON blob and exits rather than opening a stateful or request-response communication channel. This is not a functional MCP server implementation.

Tools (1)

execute_skill_actionread onlysource verified7/100

Execute a skill action (declared in mcp_server.py supported_tools)

Issues (7)

CRITICALtool ↗

No input schema visible in source code. The tool 'execute_skill_action' is mentioned in a string array but has no explicit JSON Schema definition with types, required fields, or parameter descriptions.

execute_skill_action

CRITICALtool-description ↗

Tool description is vague and circular ('Execute a skill action (declared in mcp_server.py supported_tools)'). Does not explain WHAT the tool does, WHEN to use it, or WHAT it returns. Fails pattern:tool-description.

execute_skill_action

HIGHtool ↗

Tool name 'execute_skill_action' is generic and non-specific. Uses action verbs 'execute' and 'action' which are vague. LLM cannot infer intent from the name alone. Pattern:tool recommends verb_noun pattern (get_*, create_*, update_*, delete_*, search_*, list_*) or specific action names.

execute_skill_action

CRITICALprotocol ↗

Server does not implement the MCP protocol correctly. mcp_server.py prints a single JSON handshake and exits. MCP requires a persistent request-response loop (STDIO with JSON-RPC 2.0, or HTTP with SSE). The server cannot accept tool calls or handle agent requests.

Recommendations

Rewrite the tool name to be specific and action-oriented. E.g., if the tool invokes a predefined skill by name, rename to 'invoke_skill' with parameter 'skill_name'. If it executes a generic action, split into separate tools like 'run_report', 'process_batch', etc.
Add a comprehensive tool description (50 - 200 characters): 'Invokes a pre-registered skill by name and executes it with optional parameters. Use this to trigger custom business logic or workflows defined in the server configuration.'
Define input schema with JSON Schema. Include required fields: 'skill_name' (string, enum of supported skills, e.g. ['sliding_window_trim', 'context_compression']), 'parameters' (object, optional). Each field must have a description.
Define output schema. Document what the tool returns: success/failure status, the result payload, any error codes.
Add parameter descriptions. E.g., 'skill_name: The name of the pre-registered skill to execute. Must be one of: sliding_window_trim, context_compression. See mcp_server.py for the full list.'
Implement the MCP protocol correctly. The server must: (1) use STDIO with JSON-RPC 2.0 to receive and respond to tool calls; (2) implement the 'tools/call' request handler; (3) respond with tool results in the MCP protocol format. Use a library like 'mcp' or 'anthropic-sdk' to simplify this.
Add error handling and recovery guidance. E.g., 'Skill not found. Check supported_tools in mcp_server.py.' or 'Skill execution failed: [reason]. Verify parameters match the skill's expected input schema.'

Score history

Overall score trend

First recorded score · v2 rubric

12/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	F	12	2026-07-28+	v2