An MCP server that compresses LLM prompts through semantic deduplication and filler removal to reduce token usage and API costs
This server has critical definition quality failures. The single tool 'execute_skill_action' lacks a proper input schema entirely, no parameters are defined, typed, or documented. The description is generic and does not explain what the tool actually does, when to use it, or what it returns. The mcp_server.py file contains only a stub that prints JSON and does not implement actual MCP protocol handling. Tool definitions are inferred from client.py, not explicitly registered via MCP. No output schema is documented. No error handling guidance exists. This is a non-functional MCP implementation.
Execute the LLM token context compression and cost reduction skill
No input schema defined for execute_skill_action. Tool accepts zero documented parameters despite needing raw_prompt_text and target_compression_ratio (visible in client.py).
Tool description is 67 characters and generic ('Execute the LLM token context compression and cost reduction skill'). Does not explain WHAT compression algorithm is used, WHEN to call it, WHAT the output contains, or HOW to interpret results. Fails pattern:tool-description baseline of 194 chars average and must answer: what does it do, when to use it, what does it return.
mcp_server.py is a non-functional stub. It prints a JSON status message and exits, does not implement MCP request/response handling, tool invocation, or protocol compliance. The actual compression logic lives in client.py, which is not integrated into the MCP server.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 19 | 2026-07-28+ | v2 |
No output schema documented. LLMs cannot plan downstream calls or extract fields. The client.py compress_context() method returns a dict with keys like 'compressed_prompt', 'tokens_saved_count', 'cost_savings_pct', 'semantic_similarity_score', none of these are declared in the MCP tool definition.
Tool naming 'execute_skill_action' is vague and generic. Does not start with a clear action verb (compress, reduce, optimize). 'execute_skill_action' could mean anything, LLMs cannot infer intent from the name alone. Should be 'compress_prompt_tokens' or 'reduce_context_cost'.