The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes automation scripts (launcher.sh and launcher.ps1) designed to compile the Rust-based balance lab tool using the cargo build command. These scripts subsequently execute the resulting binary to perform simulations.
[INDIRECT_PROMPT_INJECTION]: The tool's primary function involves parsing and ingesting external game data from Godot resource files (.tres) and scripts (.gd). As described in the source extraction documentation, this ingestion path represents a surface for indirect prompt injection if the project files being audited contain malicious content designed to influence the agent's balancing verdicts.
Ingestion points: Godot .tres resource files and .gd script files (specified in references/01-source-extraction.md).
Boundary markers: None explicitly implemented in the provided template logic.
Capability inventory: Execution of local build commands (cargo), execution of the compiled simulation binary, and invocation of headless Godot for calibration (specified in references/07-godot-calibration.md).
Sanitization: Not explicitly defined; the skill relies on regex patterns and deserialization for extraction.
[EXTERNAL_DOWNLOADS]: The Cargo.toml manifest file defines several external dependencies from the standard Rust registry (crates.io), including rayon for parallelism, clap for CLI argument parsing, and serde for JSON handling. These are well-known libraries required for the tool's operation.

godot-monte-carlo-balancer