bbc-skill

Warn

Audited by Gen Agent Trust Hub on Oct 2, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill accesses sensitive browser cookie databases (e.g., ~/Library/Application Support/Google/Chrome/Default/Cookies) and extracts session tokens (SESSDATA). On macOS, it attempts to fetch decryption keys from the system Keychain using the security utility.
  • [COMMAND_EXECUTION]: The skill uses subprocess.run to execute system commands including security (to access Keychain), openssl (to decrypt cookie data), and git (to perform version checks and updates).
  • [EXTERNAL_DOWNLOADS]: The skill contains logic to check for updates and perform a git pull from the author's GitHub repository (https://github.com/Agents365-ai/365-skills). While user consent is requested, this facilitates remote code updates.
  • [REMOTE_CODE_EXECUTION]: The automated update mechanism using git pull allows for the execution of remote code updates from the vendor's repository.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user-generated content from Bilibili comments (bbc/fetch.py).
  • Ingestion points: Bilibili API comment fields accessed in bbc/api.py and bbc/fetch.py.
  • Boundary markers: None identified in the data processing or storage logic.
  • Capability inventory: The skill performs network requests and local file writes to comments.jsonl and summary.json.
  • Sanitization: No sanitization or filtering of comment text is performed before saving to local storage.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 2, 2026, 01:43 AM