The Agent Skills Directory

[SAFE]: The skill provides standard development instructions for performance measurement and does not exhibit malicious behavior or security bypass patterns.
[COMMAND_EXECUTION]: The instructions include running pnpm scripts (bench, bench:compare, test) and a local Node.js utility script (benchmarking/compare/profile-diff.mjs) to analyze performance data. These are routine development tasks for the Svelte project.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read local output files such as report.txt and .md summaries to identify regressions. These ingestion points represent a data surface, but as the files are generated locally by the user's execution of project tools, the risk of malicious injection is negligible.

performance-investigation