defuddle
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill's prerequisites require the global installation of the
defuddleandjsdompackages via NPM (npm install -g defuddle jsdom). These are external dependencies that are not part of the standard agent environment. - [COMMAND_EXECUTION]: The workflow involves executing the shell command
defuddle parse "<url>"using a user-provided URL. Although the URL is wrapped in double quotes in the instructions, this does not fully mitigate command injection risks from shell expansions like backticks or dollar-parenthesis syntax. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process content from untrusted external URLs (SKILL.md). It lacks explicit boundary markers or safety instructions to the agent to ignore any malicious instructions embedded within the extracted article content. While the skill suggests sanitizing filenames, it does not specify sanitization for the extracted metadata or body text before it is displayed or saved, which could influence the agent's behavior in subsequent steps of the conversation.
Audit Metadata