anti-vibecode

Pass

Audited by Gen Agent Trust Hub on Nov 19, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a bundled Node.js script (scripts/scan.mjs) to perform static analysis on the local project directory. This script uses built-in Node.js modules (fs, path, crypto) to scan for CSS and HTML patterns without external dependencies or network activity. This is a standard and safe implementation for a project auditing tool.
  • [INDIRECT_PROMPT_INJECTION]: The skill scans and processes untrusted project source code (HTML, CSS, and Javascript) to identify design tells. This creates a surface for indirect prompt injection if a scanned file contains instructions intended to mislead the agent. However, the skill's scope is restricted to UI/UX auditing and refactoring based on the provided references/signs.md guide, and it lacks high-risk capabilities like network exfiltration or system-level privilege escalation.
  • Ingestion points: Reads files with extensions like .tsx, .html, .css, and .js within the target project directory.
  • Boundary markers: The skill does not explicitly use boundary markers (like XML tags) when processing file content, but instructions focus on pattern matching rather than interpreting logic.
  • Capability inventory: The skill uses a Node.js script for reading files and instructs the agent to edit files ('fix' mode) to update design properties.
  • Sanitization: No explicit sanitization or instruction-ignoring delimiters are applied to the analyzed project data.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 19, 2026, 01:40 AM