MCP server · trust report

io.github.Ynewtime/markitai

3 checks that ran failed — supply chain, permission scope and version behavior — so this release can’t be trusted as-is.

Publisher
io.github.Ynewtime
Repository
github.com/Ynewtime/markitai
Install
pypi:markitai, pypi:markitai
Versions
2
Inspected
5 of 6 checks · 75%
Scored
2026-09-11 · rubric 1.0.0

What we checked

Six static checks, weighted by risk. Every result reflects only what could be observed in the published package and repository — never intent.

6 checks0 clean5 flagged1 not inspected
  • Injection surface

    unscannable25% of grade

    Tool descriptions/manifest scanned for instruction-injection patterns (imperatives at the model, hidden text, 'ignore previous', data-exfil URLs).

    No tool descriptions, server instructions, prompts, or resources were fetched; nothing to scan for injection.

  • Supply chain

    fail25% of grade

    Package provenance: namespace verification, repo linkage, maintainer count, account age, postinstall scripts, typosquat distance.

    Python build script executes at build time: subprocess in build script

  • Credential hygiene

    warn15% of grade

    How the server takes secrets (env vs plaintext config vs hardcoded); secrets appearing in tool schemas.

    tests/unit/test_fetch.py: assigns a literal to api_key (test…ey (len 12)) — non-provider secret literal; prefer environment intake. tests/unit/test_llm_processor.py: assigns a literal to api_key (env:…AR (len 15)) — non-provider secret literal; prefer environment intake. tests/unit/test_llm_processor.py: assigns a literal to api_key (env:…YZ (len 32)) — non-provider secret literal; prefer environment intake. tests/unit/test_provider_discovery.py: assigns a literal to api_key (sk-n…is (len 20)) — non-provider secret literal; prefer environment intake. Reads secrets from the environment (e.g. benchmarks/llm_ab_eval.py) — the recommended intake shape.

  • Permission scope

    fail15% of grade

    Declared tools vs. breadth (filesystem, network, exec); flags shell-exec and unbounded filesystem access.

    HIGH: downloads and pipes straight to a shell (src/markitai/cli/commands/doctor.py: pipes a download into a shell). has both shell/exec and destructive filesystem writes (src/markitai/fetch_session.py: subprocess.* call; src/markitai/api.py: destructive filesystem call (shutil.rmtree/os.remove/etc.)). discloses shell/exec capability: 8 files (src/markitai/fetch_session.py, src/markitai/image.py, src/markitai/cli/main.py, +5 more) discloses broad/destructive filesystem access: 11 files (src/markitai/api.py, src/markitai/security.py, src/markitai/security.py, +8 more) Capability DISCLOSURE, not a verdict: static analysis sees the primitive is present and reachable, not whether its use is attacker-controlled.

  • Version behavior

    fail10% of grade

    Diff of tool definitions between versions; new permissions or changed descriptions in a patch release (the postmark-mcp class).

    vs prior 1.0.0 (patch bump): new capability primitive(s) in source: child_process, fs-write, http-client, net, subprocess; new hardcoded outbound host/BCC destination(s): ..., 10.0.0.1, 10.0.0.5, 10.1.2.3, 10.2.3.4, 127.0.0.1, 127.1, 169.254.169.254, 192.168.1.1, 192.168.1.50, a.b, a.com, a.example.com, a.test, alice, all, also-valid.org, api.anthropic.com, api.async.com, api.cloudflare.com, api.deepseek.com, api.example.com, api.fxtwitter.com, api.internal.corp, api.openai.com, api.twitter.com, app.corp, b.com, b.example.com, b.test, baddate.com, c.com, cached, cdn.example.com, chatgpt.com, claude.ai, code.claude.com, custom-proxy.example.com, d.com, dash.cloudflare.com, defuddle.md, developers.cloudflare.com, different, docs.anthropic.com, docs.github.com, docs.stripe.com, env-resolved-proxy.example.com, env.proxy, evil.example.com, example, example.cn, example.com, example.invalid, example.openai.azure.com, example.org, example.test, example1.com, example2.com, example3.com, exempt.example.com, expired.com, fail.com, fake.com, first, first.example, foo.bar, generativelanguage.googleapis.com, gh.io, github.com, google.com, high.com, host, httpbin.org, images.example.com, instagram.com, internal.corp, jina.ai, llm.example, localhost., low.com, lowercase, markitai.dev, medium.com, mobile.twitter.com, my-proxy.example.com, next.example, nodates.com, non-cf-site.com, nonexistent-domain-12345.invalid, nonexistent.com, nottwitter.com, ok.com, openrouter.ai, other.com, other.example, other.test, pbs.example.com, plain.example, platform.claude.com, portal.example.com, proxied.example.com, proxy, proxy.example, proxy.example.com, proxy.test, public.example, publish.twitter.com, r.jina.ai, react.dev, rebind.example, resource.openai.azure.com, schemas.openxmlformats.org, second, secure.proxy, service.corp, spa-app.example.com, test.org, twitter.com, twitter.com.fake.com, unknown.com, url1.com, url2.com, url3.com, user, valid.com, validators-regression.example.com, vimeo.com, web.proxy, wiki.corp, www.bilibili.com, www.example.com, www.instagram.com, www.twitter.com, www.w3.org, www.x.com, www.youtube.com, x.com, youtube.com.

  • Transport config

    warn10% of grade

    Remote servers: TLS and auth mode (none/token/OAuth). Local servers: whether the manifest indicates it phones home.

    sends telemetry to third-party SDK (amplitude) — disclosure, verify opt-in; manifest/source references usage stats behavior — disclosure

Version history

Each release plotted by grade against the safe line at B. A version that sinks below the line has lost its trusted standing — the shape of a rug-pull.

VersionPublishedGradeScoreInspectedChange
v1.0.1 · current2026-09-10 F37.55/6 · 75%B→F in v1.0.1
v1.0.02026-09-09 B753/6 · 55%—