unfollow-management
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill scripts interact with untrusted user data from X.com, such as usernames and follower status indicators, which constitutes an attack surface for indirect prompt injection.\n
- Ingestion points:
[data-testid="UserCell"]elements and username links (a[href^="/"][role="link"]) as described inreferences/script-details.md.\n - Boundary markers: No explicit boundary markers or instructions to ignore embedded content in user data are mentioned in the algorithms.\n
- Capability inventory: Browser console execution, DOM manipulation, state persistence via
sessionStorageandlocalStorage, and local file downloads (JSON/TXT exports).\n - Sanitization: No explicit sanitization or filtering of the ingested web content (e.g., usernames) is described.
Audit Metadata