The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes a native layout engine binary through the scripts/phoenix-diagram wrapper. This execution is tightly controlled by a launcher that validates environment variables and project paths.
[EXTERNAL_DOWNLOADS]: The scripts/phoenix-bootstrap.cjs component implements a secure download manager that fetches pre-built engine binaries from specified URLs. The manager enforces HTTPS and handles proxies and redirects securely.
[REMOTE_CODE_EXECUTION]: The skill automates the download and execution of an external binary. This process is protected by a multi-layered verification system that checks the binary's SHA256 hash, file size, and executable permissions against a signed manifest (ENGINE.json) before launching.
[DYNAMIC_EXECUTION]: The delivery process involves generating a JSON review record using a Node.js script provided in the documentation. This script is a template used to aggregate automated test results and does not process external or untrusted code.
[INDIRECT_PROMPT_INJECTION]: The skill processes diagram specifications in JSON format. While these labels could theoretically contain injection attempts, the risk is mitigated by the native Rust-based layout engine and the requirement for actual image review by a human after generation.

phoenix-diagram