promote-on-x
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data by scanning external posts and threads on X to determine relevance and craft replies. Evidence: Phase 2 (Opportunity Discovery) in SKILL.md and 'Reading Post Details' in references/X_INTERFACE.md. Ingestion points include main feeds, lists, communities, and reply threads. Capabilities include browser interaction (clicking, navigating) and writing content back to the platform. Sanitization/Mitigation: The skill includes a 'Review' phase (Phase 4) using a separate reviewer agent and a mandatory 'Approval Plan' (Phase 5) using the ExitPlanMode tool, ensuring the user reviews and approves all drafted content before it is posted.
- [DYNAMIC_EXECUTION]: The skill utilizes JavaScript injection to perform text entry into X's reply editor. Evidence: references/X_INTERFACE.md ('Typing Replies — JS Injection'). The skill uses the 'document.execCommand('insertText')' pattern to bypass X's global keyboard shortcut handlers which would otherwise intercept normal typing. This is a targeted interaction pattern for the specific platform and does not execute arbitrary code from untrusted sources.
Audit Metadata