sh.clearfront/clearfront

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.

• 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: network call 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/test_github_code_search.py: embeds a PRIVATE KEY block. (in an example/sample file — lower severity, but a real key here still leaks.) Reads secrets from the environment (e.g. clearfront/agent.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 (clearfront/agent.py: pipes a download into a shell). 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 2.7.2 (patch bump): new capability primitive(s) in source: fs-write, http-client; new hardcoded outbound host/BCC destination(s): ..., 10.0.0.1, 10.0.0.9, 127.0.0.1, 169.254.169.254, 192.168.1.1, a.com, account.shodan.io, account.shodan.io., api.abuseipdb.com, api.brightdata.com, api.github.com, api.greynoise.io, api.ip2location.io, api.openai.com, b.com, bing.com, blockstream.info, cavalier.hudsonrock.com, chess.com, cool.example, creativecommons.org, crt.sh, emailrep.io, env-server, ethereum-rpc.publicnode.com, etherscan.io, evil.com, evil.example.com, example.com, fake.com, flag-server, get.brightdata.com, github.com, google.com, google.serper.dev, graphml.graphdrawing.org, gravatar.com, haveibeenpwned.com, instagram.com, ipinfo.io, janedoe.com, linkedin.com, maps.google.com, my-server, nowhere, ollama.com, open.spotify.com, paste.ee, pastebin.com, platform.censys.io, primary-link.com, reddit.com, serper.dev, serper.dev., soundcloud.com, tiktok.com, twitter.com, url, web.archive.org, wordpress.com, www.abuseipdb.com, www.chess.com, www.google.com, www.ip2location.io, www.reddit.com, www.tiktok.com, www.twitter.com, www.virustotal.com, www.w3.org, www.youtube.com, x.com.

• Transport config

  fail10% of grade

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

  source performs runtime self-update / download-and-execute (bypasses version_behavior)