modern-web-guidance

Fail

Audited by Gen Agent Trust Hub on Sep 28, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill relies on npx -y modern-web-guidance@latest to search for and retrieve documentation. This pattern executes remote code at runtime from a public registry. The -y flag explicitly bypasses the standard confirmation prompt that allows users to verify a package before it is installed and executed, while the @latest version tag makes the skill vulnerable to malicious supply-chain updates.
  • [COMMAND_EXECUTION]: The Dynamic retrieval workflow in SKILL.md defines shell commands that take user-provided strings as arguments (search "" and retrieve ""). This creates a direct path for command injection if the agent interpolates unvalidated user input into the shell environment.
  • [INDIRECT_PROMPT_INJECTION]: The skill has an indirect prompt injection surface through its search and retrieval mechanism. 1. Ingestion points: Untrusted data enters the context via the output of the npx search and npx retrieve commands in SKILL.md. 2. Boundary markers: No specific delimiters or instructions are provided to the agent to treat retrieved content as data rather than instructions. 3. Capability inventory: The agent has shell execution capabilities (npx). 4. Sanitization: There is no evidence of sanitization or validation of the content retrieved from the external package.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Sep 28, 2026, 05:29 PM