web-auth
Pass
Audited by Gen Agent Trust Hub on Mar 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill executes the
nodecommand to run a script at/Users/avifen/.agentsys/plugins/web-ctl/scripts/web-ctl.js. This is a vendor-related tool for the 'agent-sh' plugin ecosystem, though it uses a hardcoded absolute user path.\n- [PROMPT_INJECTION]: The skill processes untrusted data from web pages, creating a surface for indirect prompt injection. \n - Ingestion points: Web content captured during authentication and verification steps in
SKILL.md. \n - Boundary markers: Present; the skill uses
[PAGE_CONTENT: ...]delimiters to separate external content. \n - Capability inventory: Shell command execution via the
nodescript. \n - Sanitization: Includes explicit safety instructions in the prompt injection warning section to ignore embedded commands.\n- [EXTERNAL_DOWNLOADS]: The skill references installing browser binaries and the Playwright framework via
npmandnpx, which are standard operations for browser automation tools.
Audit Metadata