convex-doctor

Warn

Audited by Gen Agent Trust Hub on Apr 18, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the agent to run npx convex-doctor@latest, which downloads and executes a package from the NPM registry. This package originates from an external source (nooesc) that is not identified as a pre-approved trusted provider.
  • [COMMAND_EXECUTION]: The workflow requires executing several shell commands, including npx convex-doctor@latest, npx convex codegen, npx tsc --noEmit, and npm run build. These commands are used to perform static analysis, generate code, and verify builds.
  • [PROMPT_INJECTION]: The skill processes output from an external tool to guide code remediation. This creates a surface where malicious or malformed tool output could potentially influence the agent's actions during the fix process.
  • Ingestion points: Standard output and findings from the convex-doctor tool.
  • Boundary markers: Not present; the instructions do not specify delimiters for tool output.
  • Capability inventory: The skill allows for broad code modifications across the convex/ directory based on tool findings.
  • Sanitization: No specific validation or sanitization of the tool findings is described before the agent applies fixes.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 18, 2026, 11:19 AM