web-scraping-hardened
Pass
Audited by Gen Agent Trust Hub on Apr 21, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The SAFETY.md file utilizes self-referential content (Category 8e) to present a safety evaluation with pre-determined pass/fail test cases, which is a technique designed to influence security analyzer assessments. Evidence Chain: 1. Ingestion points: The skill is designed to fetch and process external web data (HTML, JSON, sitemaps). 2. Boundary markers: SKILL.md includes instructions to confirm authorization and terms of service compliance. 3. Capability inventory: The skill utilizes shell commands (curl, bash), browser control (Playwright), and the apify CLI. 4. Sanitization: No explicit logic is provided to sanitize or validate the content ingested from target websites.
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands such as curl and grep to perform reconnaissance on remote servers to identify site structure and sitemaps.
- [EXTERNAL_DOWNLOADS]: The workflow involves the use of external packages and tools including crawlee, got-scraping, and the apify CLI to implement and deploy scrapers.
Audit Metadata