defuddle

Pass

Audited by Gen Agent Trust Hub on Apr 10, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructions prompt the installation of the defuddle-cli package from the NPM registry. This package is the core tool described in the skill and originates from the skill's author context.
  • [COMMAND_EXECUTION]: The skill utilizes shell commands for environment setup, tool verification, and file manipulation, including npm install, defuddle, date, and which.
  • [PROMPT_INJECTION]: The skill's primary function is to fetch and process content from external web URLs. This creates a surface for indirect prompt injection where malicious instructions embedded in web pages could be ingested and acted upon by the agent. \n
  • Ingestion points: SKILL.md instructions for defuddle <URL> and local HTML processing. \n
  • Boundary markers: No explicit boundary markers or instructions to ignore embedded prompts are provided in the skill documentation. \n
  • Capability inventory: The agent can execute shell commands (defuddle, npm) and manipulate files, which are capabilities that could be targeted via indirect injection. \n
  • Sanitization: While the tool cleans structural 'clutter' such as HTML boilerplate, it does not implement semantic sanitization to filter out malicious natural language instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 10, 2026, 03:04 PM