x-report
Pass
Audited by Gen Agent Trust Hub on Jul 5, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted data from X (tweets, replies, and profiles) which could contain adversarial instructions. This is a known risk for agents interacting with web content but is mitigated here by procedural human-in-the-loop safeguards.\n
- Ingestion points: Untrusted data enters the agent context via
mcp__chrome-devtools__take_snapshotas part of the analysis workflow inreferences/report-flow.md.\n - Boundary markers: The instructions mandate that the agent must 'obtain user approval' and 'confirm reporting category with the user' before any critical tool execution (e.g., clicking the 'Report' button).\n
- Capability inventory: The skill uses the
chrome-devtoolsMCP for browser navigation and interaction.\n - Sanitization: Content is analyzed directly as displayed on the platform without additional automated sanitization filters; safety relies on the required user confirmation step.\n- [SAFE]: The skill logic does not contain any obfuscated code, hidden URLs, hardcoded credentials, or unauthorized network operations. All browser interactions are consistent with the stated purpose of social media moderation.
Audit Metadata