web-content-reader

Pass

Audited by Gen Agent Trust Hub on Mar 4, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is vulnerable to Indirect Prompt Injection because it ingests untrusted data from external websites and presents it to the agent. * Ingestion points: The 'url' parameter in 'tool.js' is used to fetch content from the web via 'lib/fetcher.js' and 'lib/renderer.js'. * Boundary markers: Absent. The extracted content is returned within a 'result' object but lacks delimiters or instructions telling the agent to ignore embedded commands. * Capability inventory: The skill uses standard fetch for HTTP requests and 'playwright-core' for full browser rendering. * Sanitization: 'lib/extractor.js' uses 'cheerio' to remove script and style tags, but it does not filter or sanitize the text content for malicious instructions.
  • [COMMAND_EXECUTION]: Potential for Server-Side Request Forgery (SSRF). The tool takes a user-supplied 'url' and fetches it without restriction, which could allow an attacker to probe internal network services or access cloud environment metadata endpoints.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 4, 2026, 03:58 PM