Sealmark
Automated re-test on 2026-11-22: not checked by hand, and not yet reviewed by the maintainers.

Results / overwing-mcp

overwing-mcp

version tested 0.16.1npmtested 2026-11-22Sealmark 1.0.1methodology 0.3-draftSTANDARD run45 tools

Sealmark's automated test of overwing-mcp 0.16.1 on 2026-11-22 gave it grade A, its lowest test-area grade (protocol A and effects A). Only 4 of 45 tools could be called with default settings, so the deeper test areas cover only those tools. Of its 10 checks, 5 passed, 2 inconclusive and 3 not applicable. 3 more checks are advisory, and advisory checks never change a grade: 2 failed and 1 not applicable. The run used Sealmark 1.0.1 with the standard settings.

A

Grade A

The server's grade is its lowest test-area grade. How grading works

Only 4 of 45 tools could be called with default settings, so the deeper test areas cover only those tools.

Grade by test area

A
Protocol4/4 points
A
Effects1/1 points

Protocol checks

CheckVerdictWhat happened
protocol
Valid MCP handshake and tool schemas
pass
handshake ok, 45 well-formed tool(s)
Evidence for protocol: 1 line
  1. Probe initialize + tools/list
    Outcome server overwing 0.16.1 answered; 45 tool(s)
    Decided from the MCP exchange (protocol)
tool-poisoningcritical
No tool-poisoning indicators in tool metadata
pass
scanned 45 tool definition(s), the server instructions and serverInfo, 1 resource(s), including parameter text: nothing suspicious
robust-input
Survives malformed tool input
pass
92 invalid call(s) rejected cleanly
Evidence for robust-input: 1 line
  1. Probe 92 invalid call(s) across 19 tool(s)
    Outcome all rejected cleanly and the server stayed responsive
    Decided from the MCP exchange (protocol)
secret-echocritical
No environment secret echoed in responses
pass
canary secrets absent from tools/list and 92 captured response(s), 0 of them successful result(s)
instructions-hygiene (advisory)
Server instructions name only tools that exist
fail
the server instructions names 1 tool(s) the server does not offer (recommended_action)
refusal-quality (advisory)
Refusals say what to do next
fail
0 of 1 refusal(s) say what to do instead or why retrying will not help (0 %)

Network checks

CheckVerdictWhat happened
net-loopback-ssrfcritical
Refuses loopback targets (SSRF)
inconclusive
the listener saw nothing, but 8 of 9 probes were not explicitly refused (8 failed without saying the target was refused), so a refusal cannot be told from a tool that cannot fetch at all (a reply that only reports a failed fetch (proxy, socket, DNS) is not read as a refusal; set net.control_url to a URL the tool can reach)
net-scheme-allowlistcritical
Only http(s) schemes accepted
inconclusive
no scheme had an observable effect, but 6 of 6 scored probes were not explicitly refused, so a refusal cannot be told from a tool that cannot fetch at all (a reply that only reports a failed fetch (proxy, socket, DNS) is not read as a refusal; set net.control_url to a URL the tool can reach)
net-caps
Response size and time are capped
n/a
net-loopback-ssrf was inconclusive: the server never reached the loopback listener, so no loopback URL is known to be fetchable; response size and time caps were not measured
net-redirect
Redirects are re-validated
n/a
the server refused or never reached the loopback listener, so redirect handling could not be observed (a server that blocks loopback cannot be sent a redirect it would follow)

Effects checks

CheckVerdictWhat happened
annotation-honestycritical
Read-only tools change nothing
pass
preflight_record, preflight_overview declared read-only and changed nothing in the observed directories during a benign call (radar_status, radar_list_incidents not judged: the control call failed)
openworld-honesty
Closed-world tools stay on the machine
n/a
no tool declares openWorldHint: false that could be given a harness-owned control argument
undeclared-writes (advisory)
Unannotated tools leave the sandbox unchanged
n/a
no unannotated read-style tool could be given a harness-owned control argument

Each check id links to what it tests and how to fix a failure. The same text is in sealmark explain <id>.

Under a check, Evidence shows what it probed (sent or inspected) and what it observed in this run: up to 6 lines per check, each cut at 240 characters, for checks that are not security checks. The text is shortened, and paths and anything that looks like a secret or token are replaced. Run sealmark run --format json yourself for every line in full.

By tool

Which checks called each tool with a valid call in this run, and which did not exercise it. A check can leave a tool out for several reasons: a limit Sealmark sets, a tool left out by design, input the run could not fill in, or a call that returned an error. The CLI report gives the reason. A verdict belongs to the whole check, not to one tool: a failed check that called three tools does not say which one failed. Security checks are left out of this table.

ToolCalled byNot exercised by
evaluateNo check listed here called it with a valid call.None
evaluate_batchNo check listed here called it with a valid call.None
list_rule_setsNo check listed here called it with a valid call.None
get_rule_setNo check listed here called it with a valid call.None
create_rule_setNo check listed here called it with a valid call.None
get_evaluationNo check listed here called it with a valid call.None
list_evaluationsNo check listed here called it with a valid call.None
get_usageNo check listed here called it with a valid call.None
whoamiNo check listed here called it with a valid call.
create_accountNo check listed here called it with a valid call.None
prove_domainNo check listed here called it with a valid call.None
verify_domainNo check listed here called it with a valid call.
list_plansNo check listed here called it with a valid call.
atlas_lookupNo check listed here called it with a valid call.None
atlas_agentsNo check listed here called it with a valid call.None
atlas_register_agentNo check listed here called it with a valid call.None
atlas_verify_registrationNo check listed here called it with a valid call.None
atlas_list_registrationsNo check listed here called it with a valid call.
atlas_withdraw_registrationNo check listed here called it with a valid call.None
atlas_summaryNo check listed here called it with a valid call.
beacon_startNo check listed here called it with a valid call.
beacon_reportNo check listed here called it with a valid call.None
feedbackNo check listed here called it with a valid call.None
preflight_paymentNo check listed here called it with a valid call.None
beacon_sampleNo check listed here called it with a valid call.
radar_statusNo check listed here called it with a valid call.
radar_list_incidentsNo check listed here called it with a valid call.
radar_get_incidentNo check listed here called it with a valid call.
radar_mint_decoyNo check listed here called it with a valid call.
preflight_checkNo check listed here called it with a valid call.
preflight_verdictNo check listed here called it with a valid call.
preflight_reportNo check listed here called it with a valid call.
preflight_recordNo check listed here called it with a valid call.
preflight_overviewNo check listed here called it with a valid call.
tower_load_templateNo check listed here called it with a valid call.
tower_create_agentNo check listed here called it with a valid call.
tower_list_agentsNo check listed here called it with a valid call.
tower_revoke_agentNo check listed here called it with a valid call.
tower_capabilitiesNo check listed here called it with a valid call.
tower_decideNo check listed here called it with a valid call.
tower_submit_actionNo check listed here called it with a valid call.
tower_get_actionNo check listed here called it with a valid call.
tower_compensateNo check listed here called it with a valid call.
tower_get_receiptNo check listed here called it with a valid call.
tower_verify_receiptsNo check listed here called it with a valid call.

What was not covered

Grade over time

Since the run on

  • Server version changed from 0.16.0 to 0.16.1.

Unchanged: the same grade in every test area, the same tools with the same input schemas, the same Sealmark version (1.0.1).

  1. Av0.15.0Sealmark 1.0.0
  2. Av0.15.0Sealmark 1.0.1
  3. Av0.15.0Sealmark 1.0.1
  4. Av0.16.0Sealmark 1.0.1
  5. Av0.16.1Sealmark 1.0.1

Package facts

License
MIT
Latest release
0.16.1, published 22 Nov 2026 (the version graded on this page)
Repository
frod27/overwing-mcp
Last push
21 Nov 2026
Open issues and pull requests
3

From npm and GitHub, checked 22 Nov 2026. Plain facts about the package; they play no part in the grade.

Add it to an MCP client

This starts version 0.16.1, the version graded on this page, with the start command from the test run. For a client that reads an mcpServers JSON file:

{
  "mcpServers": {
    "overwing-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "overwing-mcp@0.16.1"
      ]
    }
  }
}

Or add it to Claude Code:

claude mcp add overwing-mcp -- npx -y "overwing-mcp@0.16.1"

The test run started the server with no API keys or other secrets, and with a cleaned environment. If the server needs some, its own documentation says how to pass them.

Run it yourself

This repeats the run with the same Sealmark version and the standard settings, starting the server the same way:

npx -y sealmark@1.0.1 run --start "npx -y overwing-mcp@0.16.1"

To confirm your run used the standard settings, add --format json --out report.json to the command. In the report, config.standard should be true and config.hash should start with sha256:d13e7893a0ac, as they did in this run. If either differs, so did the settings, and the verdicts may too (STANDARD and CUSTOM runs).

To keep the setup, put the start command in sealmark.yaml as the server.

Grade alerts

One email when the grade of overwing-mcp changes after a re-test. Nothing else, and every email has a link to stop them.

Or follow it in a feed reader: the Atom feed lists each move of the overall grade that came with a different version of this server, tested with the same Sealmark version as the graded run before.

Each of 613 verdicts was checked independently, and Sealmark was wrong on some: how reliable Sealmark is.

Is this wrong?

Reply on the GitHub issue or email disclosure@sealmark.dev and we will recheck it. Maintainers can ask for their server to be removed.

If we were wrong, we correct the page and add your reply. Security findings go to the maintainer privately first; see the disclosure policy.