Hle Benchmark Evolver

This skill appears to implement HLE report ingestion and curriculum generation, but take these precautions before installing or running it: - Ensure the expected sibling modules exist: capability-evolver (or feishu-evolver-wrapper). Inspect their src/gep/benchmarkReward.js and index.js to confirm what state files and side effects they perform. - Avoid passing untrusted commands to --eval_cmd. The pipeline will run that command via the shell (and may execute it as a temporary script using a login shell), giving it full access to the agent's environment and filesystem; it can read env vars and files or exfiltrate data. - Run the skill in an isolated/test environment first (no secrets in environment, limited filesystem access) and try it with the provided sample report to observe behaviour. - Review where the benchmark state is stored (reward.getStatePath()) and ensure you are comfortable with reads/writes to that path. - If you must run eval_cmd, prefer to run a controlled evaluator executable you trust and pass a restricted environment (or run inside a sandbox/container). If you want a safer install, ask the skill author to: declare the dependency on capability-evolver, document the state path and files touched, and add explicit warnings and safeguards around executing arbitrary eval_cmd shell commands.