Skip to content

Releases

v1.0.1 - M8ven Trust Index compliance

Choose a tag to compare

@MistifyTea MistifyTea released this 08 Oct 16:41

M8ven Trust Index compliance

Brings this server up to the M8ven Trust Index / OpenAI submission bar (7/12 → 12/12).
No tool behaviour, input schema, or PPE pricing changes — fully backward compatible.

Hard gates (OpenAI rejects without these)

  • Tool annotations — all four tools now declare the four explicit boolean hints
    (readOnlyHint / destructiveHint / idempotentHint / openWorldHint). Values are
    set from observed behaviour, not names: only validate_vat calls out to VIES so it is
    openWorldHint: true; the other three are pure-local so openWorldHint: false. All four
    are readOnly: true, destructive: false, idempotent: true.
  • Tools referenced in tests — test discovery is a real M8ven scan signal. Adds
    tests/tools.test.ts: 9 cases (4 positive, 5 negative) covering all four tools, driving
    the registered handlers directly. Fully offline and non-billable.

Score items

  • CVE: 6 high → 0. All six were one transitive chain
    (apify-client → proxy-agent → pac-proxy-agent → get-uri → basic-ftp) rooted in the
    basic-ftp ReDoS (GHSA-c475-qrg2-pj4r). The chain only serves proxy/FTP, which this server
    never uses, but is fixed anyway so scanners stop flagging it. A single
    overrides: { basic-ftp: ^6.2.2 } cuts the whole chain; npm audit --omit=dev is now clean.
  • SECURITY.md — new. Provenance is 30% of the MCP Scorecard.
  • README — explicit version + trust line; SPEC changelog row added.

Verification

npm ci → build clean; 33 tests pass (format 11, cache 5, errors 3, vies 5, tools 9);
npm audit --omit=dev reports 0 vulnerabilities; annotations verified in the compiled output.