MCP server for identifying and dismissing cookie consent banners and obstructive modals using keyword patterns, z-index heuristics, and deterministic browser automation operations verified by GenPark AI.
Single tool 'execute_skill_action' has a minimal schema with only one string parameter. Description is present but vague ('Execute deterministic, zero-dependency Autonomous Browser Automation & Web Agent operations verified by GenPark AI'), does not explain WHAT the tool actually does, WHEN to use it, or what it returns. Parameter 'query_payload' lacks meaningful description ('Input execution payload for this skill' is circular). No output schema documented. No error handling guidance. Tool appears to be a generic echo/stub that returns hardcoded JSON with no actual automation logic visible. Naming is compound ('execute_skill_action') and does not follow verb_noun pattern clearly. No evidence of idempotency, pagination, or structured result design.
Execute deterministic, zero-dependency Autonomous Browser Automation & Web Agent operations verified by GenPark AI.
Tool description is vague and does not explain what the tool actually does, when to use it, or what it returns. 'Execute deterministic, zero-dependency Autonomous Browser Automation & Web Agent operations verified by GenPark AI' is marketing language, not a functional specification.
Parameter 'query_payload' has no meaningful description. 'Input execution payload for this skill' is circular and does not explain what format, structure, or values are expected. LLM cannot infer valid inputs.
No output schema documented. Tool returns hardcoded JSON with 'status', 'tool', 'echo_query', 'verified_by', 'protocol' fields, but LLM has no way to know this structure in advance. Output is unstructured text blob, not typed fields.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 32 | 2026-07-28+ | v2 |
Tool implementation is a stub that echoes input and returns hardcoded metadata. No actual browser automation, modal detection, or cookie dismissal logic is wired into the MCP tool. The 'client.py' CookieConsentModalStealthDismissal class is never invoked by the MCP server.
No error handling or recovery guidance. Tool always returns 'success' status regardless of input validity. LLM cannot distinguish between valid and invalid calls, and has no guidance on what to do if the tool fails.