data-scraper-agent

Fail

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: CRITICALINDIRECT_PROMPT_INJECTIONPRIVILEGE_ESCALATIONCOMMAND_EXECUTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external web sources like HTML, JSON, RSS feeds, and PDF documents using requests, playwright, and BeautifulSoup as seen in SKILL.md and payloads.md. This creates an attack surface for indirect prompt injection where malicious instructions in external content could influence the agent. The skill lacks explicit boundary markers or sanitization logic. Capabilities include file system writes, network requests, and subprocess execution.
  • [PRIVILEGE_ESCALATION]: The payloads.md file contains a command to manage the Tor service which requires elevated permissions: sudo systemctl start tor.
  • [COMMAND_EXECUTION]: The skill interacts with the searchsploit CLI tool using Python's subprocess module and demonstrates the use of parallel for batch processing shell commands in payloads.md and the rate-limiting guide.
  • [DATA_EXFILTRATION]: Documentation in payloads.md provides patterns for searching and extracting sensitive credentials like API keys and private keys from GitHub repositories using the gh CLI tool.
  • [EXTERNAL_DOWNLOADS]: The skill fetches vulnerability and advisory data from external platforms including services.nvd.nist.gov, cveawg.mitre.org, and api.github.com. It also provides examples for enumerating hidden service directories on the Tor network in payloads.md.
Recommendations
  • Contains 1 malicious URL(s) - DO NOT USE
Audit Metadata
Risk Level
CRITICAL
Analyzed
Sep 19, 2026, 04:36 PM