flashdata-mcp

Pass

Audited by Gen Agent Trust Hub on Nov 3, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data retrieved from web searches and YouTube transcripts, creating a surface for indirect prompt injection.
  • Ingestion points: Data enters the context via google_search, youtube_search, and youtube_transcript tools as described in SKILL.md and references/video-research.md.
  • Boundary markers: SKILL.md contains explicit instructions to treat all retrieved content as source data and to disregard any embedded instructions that attempt to change account settings, execute code, or disclose credentials.
  • Capability inventory: The skill is scoped to FlashData MCP tools for data retrieval (search, metadata, transcripts) and budget management.
  • Sanitization: The skill relies on natural language instructions to the agent to distinguish between evidence and synthesis, rather than programmatic sanitization.
  • [EXTERNAL_DOWNLOADS]: The documentation refers to external resources for local installation and configuration.
  • Evidence: references/connection.md mentions the GitHub repository github.com/flashdatadev/flashdata-mcp and the npm package @flashdata-dev/mcp.
  • Context: These resources originate from the skill's official author and are standard for configuring the required Model Context Protocol (MCP) server environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 3, 2026, 06:39 PM