last30days
Audited by Socket on Sep 15, 2026
7 alerts found:
Anomalyx3Securityx4The code is primarily a setup and authentication helper. It contains no clear malware, obfuscation, destructive behavior, or command injection. The main security concern is intentional transmission of a locally obtained GitHub CLI token to api.scrapecreators.com, plus sensitive browser-cookie discovery and automatic yt-dlp installation. These actions may be legitimate for the application, but they require explicit user consent, trust in the remote service, and careful credential handling. Review the service ownership and privacy policy before use.
The fragment is a browser-cookie extraction utility. It does not itself show exfiltration, persistence, command execution, or obfuscation, but it intentionally reads highly sensitive authentication cookies, including Windows Firefox cookies through WSL, and returns them to its caller. This is a significant credential-access capability; whether it is malicious depends on the surrounding application's authorization and use. The fragment is incomplete as supplied, and delegated macOS extractor behavior is unavailable.
The code is a browser-cookie extraction utility for macOS. It obtains browser encryption material from the Keychain and returns selected plaintext Chrome or Brave cookies, including potentially reusable authentication sessions. That is inherently high-impact credential-access behavior and should only be used with explicit authorization. No exfiltration or other malicious payload is shown in this fragment, so malicious intent cannot be established from the file alone.
The script is an automation wrapper for an A/B research test, not intrinsically malicious. However, it launches two externally defined Claude Code workflows with permission checks disabled and embeds untrusted user input in their prompts. This creates a meaningful command-execution and prompt-injection risk if the skills, project instructions, or topic content are compromised. The script should avoid --dangerously-skip-permissions, validate the topic and output directory more strictly, and preserve execution errors for review.
This module is a high-sensitivity credential-handling component: it specifically reads Twitter/X session cookies (auth_token and ct0) from environment variables or from local browser cookie stores (via a third-party helper) and then constructs and returns an HTTP Cookie header containing those secrets. While it shows no direct exfiltration or malicious payload execution within this snippet, the capability to harvest authentication material is inherently dangerous if misused or if the dependency/downstream request handling is compromised. The dynamic import of @steipete/sweet-cookie adds supply-chain execution risk. Review downstream usage of the returned cookieHeader and ensure strict protection of logs, telemetry, and network destinations.
This module is a macOS-specific parser that reads Safari’s local Cookies.binarycookies file and extracts selected cookie values for a provided domain and cookie name set. It contains no direct network exfiltration, command execution, or obfuscation in this fragment. The security risk is primarily its capability to harvest highly sensitive authentication/session cookies, which could be misused by other parts of a containing package. Additionally, the snippet shows a likely defect (`return resul`), which may impair execution but does not negate the sensitive extraction design.
No clear malware behaviors are evident in this fragment (no eval/dynamic execution, no local persistence, and no explicit data exfiltration beyond normal HTTP requests). However, there is a significant security anomaly: getBaseHeaders() hardcodes an Authorization Bearer token in addition to accepting/storing user-provided auth cookies and CSRF token, which materially increases credential exposure and supply-chain risk. The module also suppresses errors during query-id refresh, which can reduce operational transparency. Because the snippet is incomplete beyond ensureClientUserId(), full determination of destinations and how responses are handled is not possible here.