The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructions require the agent to download and execute an installer via curl -sSL ... | sh from the URL https://raw.githubusercontent.com/purong-huang-1121/skills-store/main/install.sh. This provides the repository owner with arbitrary command execution on the user's system.
[COMMAND_EXECUTION]: The installer is executed every time the CLI is not found and is checked for updates every 12 hours. This creates a persistent lifecycle for the remote code execution.
[CREDENTIALS_UNSAFE]: The skill explicitly asks users to store their EVM_PRIVATE_KEY in a .env file for trading functions. Combined with the execution of untrusted remote code, this creates a critical risk of private key theft and loss of funds.
[PROMPT_INJECTION]: The skill lacks boundary markers when processing untrusted market data from external prediction market searches.
Ingestion points: plugin-store polymarket search and markets results.
Boundary markers: Absent.
Capability inventory: Local file access (.env), network access via plugin-store, and transaction signing capabilities.
Sanitization: No sanitization of market names or question text is specified before processing.

dapp-polymarket