scrapfly-agent-rules
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates a web-scraping agent which ingests untrusted data from arbitrary websites, creating a potential surface for indirect prompt injection.
- Ingestion points: Tools such as web_scrape, cloud_browser_open, and take_snapshot (documented in SKILL.md) return external web content.
- Boundary markers: Not explicitly defined in this ruleset.
- Capability inventory: The agent is instructed to perform browser actions (click, fill) and create monitoring alerts.
- Sanitization: Not explicitly defined; however, Rule 8 provides a significant manual mitigation by requiring human review and dry-runs before committing alerts.
- [SAFE]: No malicious code, obfuscation, or unauthorized data access patterns were identified. External URLs are directed to the vendor's official infrastructure (scrapfly.io) and example configuration for API keys uses safe placeholders.
Audit Metadata