The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input (blockchain addresses, token values, RPC URLs) through an interactive bulk form. These values are stored in configuration objects and interpolated into shell commands.
Ingestion points: User responses collected via the AskUserQuestion tool in SKILL.md.
Boundary markers: The skill does not define specific prompt delimiters, but it enforces rigorous validation rules for every collected parameter.
Capability inventory: The skill utilizes Bash (via curl), Write (to save configuration files), and specialized MCP tools for supply schedule generation.
Sanitization: The skill implements mandatory regex validation for Ethereum addresses (^0x[a-fA-F0-9]{40}$) and numeric values, along with a strict blacklist for shell metacharacters and whitespace (;, |, &, $, backticks, etc.) to prevent command injection during interpolation.
[EXTERNAL_DOWNLOADS]: The skill uses curl to fetch the current block number from blockchain RPC endpoints. While it provides a list of known public RPCs, it also permits custom user-provided endpoints. The skill includes specific instructions to validate that these endpoints use the https:// scheme and are sanitized against injection attacks.
[COMMAND_EXECUTION]: The skill provides instructions for the agent to prepare and execute a local MCP server via shell commands (chmod +x setup.sh, ./setup.sh, and python3 server.py). These commands are limited to the plugin's internal directory structure for generating supply schedules.

configurator