x-article-draft-uploader

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [DATA_EXFILTRATION]: The script scripts/export_x_cookies_from_chrome.py accesses sensitive browser data by reading the Google Chrome cookie database.
  • Evidence: The script opens the SQLite database at ~/Library/Application Support/Google/Chrome/Default/Cookies to retrieve authentication tokens.
  • Context: The access is restricted to x.com and twitter.com domains. Extracted cookies are saved locally at ~/.ailu/secrets/x/cookies.json with 0600 (read/write by owner only) permissions, which is a standard security practice for handling session secrets.
  • [COMMAND_EXECUTION]: The skill utilizes system commands to access the macOS keychain and manage browser dependencies.
  • Evidence: scripts/export_x_cookies_from_chrome.py invokes /usr/bin/security via subprocess.run to obtain the 'Chrome Safe Storage' password required for cookie decryption.
  • Evidence: The documentation instructs users to execute playwright install chromium to set up the automation environment.
  • [INDIRECT_PROMPT_INJECTION]: As the skill ingests untrusted Markdown files for processing and upload, it possesses an indirect prompt injection surface.
  • Ingestion points: Markdown files provided as input to the upload_markdown_to_x_article.py script (referenced in SKILL.md).
  • Boundary markers: The skill uses specific internal tokens like X_TABLE_MARKER_* to delimit content during the transformation process.
  • Capability inventory: The skill possesses network access (via Playwright), file system read/write access (for article assets and results), and command execution capabilities.
  • Sanitization: A dedicated parser in scripts/parse_markdown.py handles the conversion of Markdown to HTML, providing a layer of sanitization against raw HTML injection.
  • [EXTERNAL_DOWNLOADS]: The skill's setup process involves downloading code and dependencies from external repositories.
  • Evidence: The README.md provides an installation command that fetches the skill from the author's GitHub repository at github.com/mcncarl/yichen-skills.
  • Evidence: The requirements.txt file specifies external Python dependencies including playwright and pycryptodome.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 05:58 AM