Advisory

MAL-2026-18426

Also known as: None

Malicious package Its source marks this package as malicious: it is not a vulnerability in a legitimate one.

Summary

Malicious code in is-it-ai (PyPI)

Details

---
_-= Per source details. Do not edit below this line.=-_

Source: amazon-inspector (6c41881d21c4cb046c011c52db3d7725f8d936685710b170f3d4c116032a5158)
The package's public verify() function POSTs a JSON payload to the hardcoded endpoint https://cdn-9zp0pi12ck0dmfig-3ozmpyh7ja-uc.a.run.app/v1/session/info containing OS details, Claude Code / Cursor / Codex environment variables (CLAUDE_PID, CLAUDE_CODE_ENTRYPOINT, CLAUDE_CODE_MANAGED_SETTINGS_PATH), and a 'thumbprint' built from MD5 hashes of resolved AI-agent marker paths (e.g. /opt/homebrew/bin/claude, CLAUDE.md, MCP gateway.cer files, managed-preferences plists). The server response is then parsed for two attacker-controlled fields: a 'report' URL and a 'lock' filesystem path. _parse_report downloads the URL via urllib.request.urlretrieve, writes the bytes to the server-specified path via os.replace, and executes it — os.startfile(file) on Windows, chmod 0700 + subprocess.Popen([file], start_new_session=True) on macOS. The remote server fully controls both the bytes written to disk and the destination path. The control-plane host is an anonymous Cloud Run default subdomain unrelated to the author identity in package metadata, consistent with an attacker-hosted collector-and-dropper rather than first-party vendor infrastructure. The AI-agent fingerprinting lets the operator select payloads for developer hosts running Claude Code, Cursor, or Codex.

Source: kam193 (6a8580750a7d490d65866e9407c451efa8dbad9a90c35a54b87416c57eef1fea)
In this campaign, specific versions of involved affected packages hide malicious actions. The malicious code is hidden in specific versions of a dependency (is-it-ai), intentionally pinned by specific versions of dependent packages. All package versions were uploaded with malicious intention; the ones without malicious references are clearly intendent only to hide a cover for the rest. When triggered, the malicious action fingerprints the environment and contacts the remote server. If it responds with an URL, it will be downloaded and executed.

---

Category: MALICIOUS - The campaign has clearly malicious intent, like infostealers.

Campaign: 2026-10-is-it-ai

Reasons (based on the campaign):

  • Downloads and executes a remote executable.
  • The malicious code is intentionally included in a dependency of the package
  • targetted-attack
  • action-hidden-in-lib-usage

Severity

Not rated by the source.

Dates

Published
2026-10-31 12:48 UTC
Modified
2026-11-12 03:45 UTC

Affected packages and fixes

is-it-ai

PyPI · All systems

Affected

  • Versions: 0.1.0, 0.1.1, 0.1.3

Fixed versions

None listed

Fix

No fixed version is listed.

Fix commands are the export's own generic commands. Ojuju never runs them. Check them against your setup first.

Check your own computer: guard packages check

View this record on OSV.dev

Source

Source: OpenSSF Malicious Packages · Apache-2.0

Sources & licences