The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill initializes a wallet that stores a sensitive private key in a local file at ~/.t2000/wallet.key. The documentation explicitly states that the key is stored in plain text (Bech32 format) without a PIN or encryption, though it uses standard file permissions (0o600) to restrict access.
[EXTERNAL_DOWNLOADS]: The skill instructs the installation of the @t2000/cli package from the public npm registry to provide core functionality.
[COMMAND_EXECUTION]: The skill relies on several shell commands to configure the environment, including npm install for software installation, t2 init for wallet creation, and t2 balance for state verification.
[INDIRECT_PROMPT_INJECTION]: By connecting to a hosted Model Context Protocol (MCP) server at https://mcp.t2000.ai/mcp, the skill introduces an external dependency that provides tools and prompts to the agent, creating a surface for indirect instruction injection.
Ingestion points: Tool definitions and prompt content retrieved dynamically from https://mcp.t2000.ai/mcp.
Boundary markers: No specific delimiters or "ignore instructions" warnings are defined for the tool outputs.
Capability inventory: The associated CLI tool (t2) has the capability to transfer funds, swap tokens, and interact with on-chain services.
Sanitization: There is no mention of sanitization or filtering logic for the data returned by the hosted service.

t2000-setup