The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs users and agents to execute unverified remote code via a piped shell command: curl -sSL https://raw.githubusercontent.com/caiovicentino/polymarket-mcp-server/main/quickstart.sh | bash. This pattern is highly insecure as it executes a script from an external repository directly in the system shell without prior inspection.
[EXTERNAL_DOWNLOADS]: The installation process relies on cloning a third-party repository from GitHub (github.com/caiovicentino/polymarket-mcp-server) and executing local scripts (install.sh, quickstart.sh) which are outside the immediate package's static control.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data from prediction markets, which can be manipulated by third parties to influence the agent's behavior.
Ingestion points: Data enters the context through market discovery tools (search_markets, get_trending_markets) and detailed market lookups (get_market_details).
Boundary markers: None identified; market descriptions and titles are processed as natural language instructions for the analysis tools.
Capability inventory: The skill has high-privilege capabilities including placing limit orders (place_limit_order), market orders (place_market_order), and performing "smart trades" (smart_trade) based on processed instructions.
Sanitization: There is no evidence of sanitization or filtering for market-provided content before it is passed to AI-powered analysis tools like analyze_market_opportunity or optimize_portfolio.

polymarket-mcp-server