defuddle

Pass

Audited by Gen Agent Trust Hub on Apr 2, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes the defuddle command-line tool to parse web pages as part of its primary functionality documented in SKILL.md.\n- [EXTERNAL_DOWNLOADS]: Recommends the installation of the defuddle NPM package, which is a utility provided by the author kepano.\n- [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection.\n
  • Ingestion points: Processes web content from external URLs using the defuddle parse <url> command in SKILL.md.\n
  • Boundary markers: Absent; instructions do not specify the use of delimiters or warnings to ignore instructions within fetched content.\n
  • Capability inventory: The skill uses shell command execution to process data.\n
  • Sanitization: Absent; content is extracted and converted to markdown without explicit sanitization or instruction filtering.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 2, 2026, 11:00 PM