NYC

vercel-blob

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFE
Full Analysis
  • Prompt Injection (SAFE): No instructions were found that attempt to override agent behavior or bypass safety constraints. The content is strictly technical documentation and boilerplate.
  • Data Exposure & Exfiltration (SAFE): The skill mentions the requirement of environment variables (BLOB_READ_WRITE_TOKEN) but does not contain hardcoded credentials or scripts that exfiltrate local data to unauthorized domains.
  • Obfuscation (SAFE): No evidence of Base64, zero-width characters, homoglyphs, or encoded executable commands was detected in any of the analyzed files.
  • Unverifiable Dependencies & Remote Code Execution (SAFE): The project uses official, versioned packages from trusted sources (Vercel and Meta). There are no suspicious 'curl | bash' patterns or unauthorized package installations.
  • Privilege Escalation (SAFE): No commands requiring elevated privileges (sudo) or modifications to system-level permissions were found.
  • Persistence Mechanisms (SAFE): There are no scripts that attempt to install background services, cron jobs, or modify shell profiles.
  • Metadata Poisoning (SAFE): The metadata in plugin.json and the README accurately reflect the skill's purpose without deceptive instructions.
  • Indirect Prompt Injection (LOW): While the skill is designed to handle user-provided files (images, PDFs), the risk is minimal as the logic is limited to standard storage operations using the Vercel SDK. The code snippets provided include client-side token security patterns.
  • Time-Delayed / Conditional Attacks (SAFE): No logic was found that gates execution based on dates, times, or specific environment triggers.
  • Dynamic Execution (SAFE): The skill does not generate or execute code at runtime using eval, exec, or dynamic library loading.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:02 PM