entroly

Entroly LobeHub MCP Score Audit

This document records evidence for the LobeHub MCP score without treating an external marketplace score as a substitute for product quality.

Target listing:

https://lobehub.com/mcp/juyterman1000-entroly?activeTab=score

Verified public baseline

A GitHub-hosted audit fetched and decoded the public page on 2026-07-13. The listing reported:

The 45 points were deployment (15), non-manual installation (12), license detection (8), and README (10). Tools and validation are required and were the two missing required items.

Scoring contract and deduction map

LobeHub’s public implementation assigns 100 points:

Criterion Weight Required Public observation Repository evidence Classification/action
Claimed listing 4 No False; the page requests its exact GitHub badge before claim verification Exact LobeHub ownership badge is committed in the primary README Legitimate ownership proof; public claim check remains external
Non-manual deployment 12 No Passing server.json declares PyPI/uvx and npm/npx install paths No product change required
Any deployment 15 Yes Passing; two methods Two canonical package deployment options No product change required
Detected license 8 No Passing but stale as MIT Current repository and package metadata are Apache-2.0 External metadata refresh required; points already awarded
MCP prompts 8 No Zero Context-optimization and evidence-verification workflows Genuine capability added and protocol-tested
README 10 Yes Passing but stale Current primary and package READMEs External metadata refresh required; points already awarded
MCP resources 8 No Zero Bounded entroly://health and entroly://stats resources Genuine read-only capability added and protocol-tested
MCP tools 15 Yes Zero despite the existing production tool surface FastMCP tools for optimization, exact recovery, receipts, memory, and verification Stale or failed external validation; published-artifact validation required
Runtime validation 20 Yes False Stdio initialize/list/get/read/call protocol tests LobeHub must re-run validation after publication

A required criterion missing causes grade F regardless of percentage. If all required criteria pass, 80% or higher is A, 60–79% is B, and lower is F.

Release isolation

The ClawHub v1.0.54 install-metadata correction was completed in PR #124. Its merge commit has a successful ClawHub Public Listing status for entroly-openclaw v1.0.54. This LobeHub work remains isolated in PR #125 and does not rewrite the completed v1.0.54 release.

Legitimate remediation

The added MCP prompts are reusable workflows backed by real Entroly tools. The added resources expose only bounded, read-only operational summaries. They do not expose source content, receipt bodies, filesystem paths, credentials, or unbounded logs.

A clean Python 3.12 validation job installed Entroly from the branch and proved that a fresh stdio server can:

  1. initialize successfully;
  2. list and render both prompts;
  3. list entroly://health and entroly://stats;
  4. read and parse the bounded health resource;
  5. continue exposing the existing tool surface; and
  6. pass the local score-evidence regression tests.

The repository audit script separately computes local readiness and labels it as local evidence rather than the public score. With protocol validation in the same run, the repository has evidence for 96/100 before the external claim point. This is not the current public LobeHub score.

Remaining external gates

The work is not complete until all of these are visibly confirmed:

No release note or PR may claim an improved LobeHub score before that public confirmation.