custom-skills-threads-research

Warn

Audited by Gen Agent Trust Hub on Aug 28, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill directs the user to install the context-mode package from the NPM registry using npx. This tool is maintained by an external third-party developer.
  • [INDIRECT_PROMPT_INJECTION]: The skill indexes data from public Threads posts and Google Search results, which can contain untrusted content.
  • Ingestion points: External web content is retrieved via Playwright navigation and snapshots as described in Phase 1 and 2.
  • Capability inventory: The skill uses browser automation and a local FTS5 indexing system to store and query data.
  • Boundary markers: No specific delimiters or "ignore instructions" warnings are used when indexing the scraped Markdown content.
  • Sanitization: The skill extracts structured data using JavaScript but does not perform safety filtering on the text content before indexing.
  • [COMMAND_EXECUTION]: The skill includes instructions to execute shell commands for adding new MCP servers to the agent's configuration using the claude mcp add command.
  • [DYNAMIC_EXECUTION]: The skill utilizes the browser_evaluate tool to execute custom JavaScript code within the browser context to parse search engine result pages.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 28, 2026, 08:16 AM
Security Audit — agent-trust-hub — custom-skills-threads-research