The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses child_process.spawnSync and execFileSync within its validation and timing scripts (benchmark-contract.test.mjs and paired-timing.mjs). These are used to run legitimate internal testing and benchmarking commands (e.g., bun bench.ts, pnpm bench, node). All command inputs are internally defined or derived from local configuration files, following the project's performance methodology.
[EXTERNAL_DOWNLOADS]: The skill includes instructions and scripts for installing internal dependencies, such as pnpm bench:editor:install, which targets the project's own benchmarks/editor package. These are standard development workflows for the vendor 'udecode' and do not involve untrusted remote sources.
[DYNAMIC_EXECUTION]: The script validate-benchmark-plan.mjs dynamically reads and validates local markdown files using project-specific parsers. This is a static analysis of plan files and does not execute arbitrary code or load untrusted data.
[SAFE]: The skill provides structured templates and methodologies for performance reviews and benchmarks. It includes correctness guards and verification steps to ensure that performance optimizations do not regress application behavior. All scripts and references are focused on legitimate performance engineering practices within the 'udecode' project scope.

benchmark