Autonomous Agent Toolkit

What to check before installing: - Review the two code files locally (scripts/generate_agent.py and tests/health-check.js). The generator is benign and writes workspace files, but the health-check script has brittle and possibly unsafe assumptions: it hardcodes a Windows admin workspace path (C:\Users\Administrator\.openclaw\workspace), looks for other skills (e.g., 'self-evolve'), and expects files that the generator does not create. Running that test unmodified could read unexpected files in your environment — do not run it with elevated privileges. - Confirm which external APIs/crons you will configure. The SKILL.md shows cron examples that check sales APIs and other external systems; you will need to supply credentials for those. The skill metadata declares no env vars/keys — ensure you do not provide secrets unless you understand exactly which cron/tasks will use them and where they are stored. - Address metadata mismatches (registry version vs _meta.json) and the test's dependency on other skills. Ask the maintainer or inspect the repo to ensure you have the intended release and that tests are up-to-date. - If you plan to enable autonomous cron jobs, apply least privilege: limit what the agent can do (kill switches, red lines), run in a sandbox or isolated workspace, and avoid granting external service keys until you've audited the cron actions and code. - If unsure, run the generator in a disposable directory and inspect outputs before wiring crons or giving network/credential access. If you need higher assurance, request an updated release that removes hardcoded paths, documents required credentials, and aligns metadata/tests with the generator's behavior.