tbs-scenario-builder

This skill appears to implement the TBS scenario creation workflow, including preflight checks and actual writing to a remote TBS API — which is consistent with its description — but there are important mismatches you should consider before installing: 1) Missing credential declarations: SKILL.md and scripts expect an access token (XG_USER_TOKEN) and a TBS_BASE_URL, but the package metadata lists no required env vars. Do not install or enable this skill unless you can supply a scoped access token and you understand which token will be used. Prefer giving a least-privilege, audit-friendly token (or a read-only/dry-run token) for testing. 2) Runtime network installs: The skill instructs using 'npx' to install a required auth skill at runtime (and falls back to a GitHub URL). That means code will be fetched and executed from the network. If you plan to run this skill, review the cms-auth-skills source that will be installed and consider pre-installing it in a controlled way rather than allowing the agent to run npx automatically. 3) Side effects: The executor script will call TBS endpoints and can create master data (e.g., POST /basic/drugs) automatically. Use the provided '--dry-run' / preflight mode first and verify behavior in a non-production TBS environment. Insist on explicit user confirmation (the skill requires the exact '确认' token) before any write — but verify the agent enforces that requirement in practice. 4) Audit the code: If you plan to enable this skill in an environment with real data, review the bundled scripts (especially tbs_write_executor.py and tbs_master_data_resolve.py) so you understand exactly which HTTP calls and local writes will occur. Check logging, error handling, and whether tokens are leaked to any unexpected endpoints. 5) Operational controls: Prefer running this skill in a sandbox first; provide minimal-scope credentials; require human confirmation before persist-and-execute; consider network egress controls or allowlisting the TBS_BASE_URL to limit where credentials can be used. If these points are acceptable and you can review or provide the necessary runtime tokens and test environment, the skill can be used. If you cannot supply a scoped token or review the downstream scripts, treat this package as potentially risky and avoid enabling its automatic execution.