MCP stdio server executing WASM tools in the Ephemora Cell. Fast, capability-based WASM execution with explicit CPU, memory, time, I/O, and filesystem limits.
Three tools are defined, but definition quality is poor across the board. All three tools have minimal or missing parameter descriptions, no proper input validation guidance, and lack output schema documentation. The 'echo' tool has a completely open schema (additionalProperties: true) with zero parameter documentation. The 'clock' tool has a properly constrained schema but lacks any parameter description. The 'get-policy' tool has a single optional parameter with minimal guidance. None of the tool descriptions explain what output structure is returned, breaking downstream tool composition. Error handling is not addressed. The server appears to be a minimal WASM sandbox executor rather than a production-grade tool gateway, and tool definitions reflect this prototype-level quality.
Returns the current UTC time in ISO-8601 format and Unix milliseconds
Echoes its arguments back
Returns the effective sandbox policy for Cell tools ("Verified. Not claimed."): fuel budget, memory limit, threads, preopens (configured), network policy and the security baseline incl. wasmtime version. Pass {"tool": "<name>"} for one tool or no arguments for the whole registry.
echo tool has completely unconstrained schema (additionalProperties: true) with zero parameter descriptions. LLMs cannot determine valid inputs or expected fields.
No tool describes its output schema or structure. Downstream tools and LLMs cannot infer what fields are returned, breaking tool chaining and composition.
clock and get-policy parameter descriptions are present but minimal. 'Optional tool name to get policy for a specific tool' lacks guidance on expected format (name vs ID vs pattern), and clock parameters are completely undocumented despite accepting no parameters.
No error handling guidance. If a tool fails, LLMs have no direction on what to do next, how to retry, or whether the failure is recoverable.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 50 | 2026-07-28+ | v2 |
echo tool naming is vague. 'echo' does not follow verb_noun convention (e.g., 'return_input', 'reflect_params'). The name alone does not convey the tool's purpose clearly to LLMs.
get-policy tool name begins with a verb but lacks clarity. Does 'policy' mean security policy, rate limit policy, or feature access policy? A verb_resource_aspect name like 'get_sandbox_policy' would be clearer.