article-extractor
Pass
Audited by Gen Agent Trust Hub on Mar 25, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests data from arbitrary URLs and displays a preview in the agent context without sanitization.\n
- Ingestion points: External URLs processed by curl, reader, or trafilatura in SKILL.md.\n
- Boundary markers: Absent; no clear separation between web data and agent instructions.\n
- Capability inventory: Bash tool for system commands and Write tool for file creation.\n
- Sanitization: None identified for embedded instructional content.\n- [COMMAND_EXECUTION]: Executes shell commands to install dependencies and run data processing pipelines.\n
- Logic: Dynamic execution of an inline Python HTML parser using python3 -c with input from stdin.\n
- Risk: Command construction uses variables derived from external web content (titles).\n- [EXTERNAL_DOWNLOADS]: Fetches content from user-specified URLs and installs third-party tools from official package managers.\n
- Tools: reader-cli, @mozilla/readability-cli, and trafilatura.
Audit Metadata