Skills
Navigate
skills/unbrowse-ai/sites/buffer/Gen Agent Trust Hub

buffer

Pass

Audited by Gen Agent Trust Hub on Sep 28, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill fetches content from external web pages (buffer.com), which constitutes an indirect prompt injection attack surface.\n
  • Ingestion points: buffer_com__read_page and buffer_com__get_resources_guides_courses in SKILL.md fetch content from the web.\n
  • Boundary markers: No specific delimiters or instructions to ignore embedded commands are provided.\n
  • Capability inventory: The skill can perform actions on the site (unbrowse.run, unbrowse.browse.act) based on agent decisions.\n
  • Sanitization: No explicit filtering of fetched content is described.\n- [EXTERNAL_DOWNLOADS]: The skill connects to remote vendor-owned resources for tool execution and API access.\n
  • Evidence: References to https://unbrowse.ai/mcp/buffer.com and https://unbrowse.ai/api/v1/... in SKILL.md and server.json.\n- [DYNAMIC_EXECUTION]: The skill provides a mechanism to record actions and generate new tools on the vendor's platform.\n
  • Evidence: SKILL.md describes tools like unbrowse.browse.open that "compile" new tools for the MCP server.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 28, 2026, 10:56 AM