add-tests

The skill's intent (produce project-conformant tests and verify they pass) is legitimate and useful. However, implementing it as documented — which instructs the agent to run repository test suites and invoke tooling like npx — presents moderate supply-chain and data-exposure risks when used against unvetted repositories. Key risks: execution of arbitrary repository code (test suites and hooks), potential exfiltration of environment or filesystem secrets via test output, and transient dependency installation via npx. Mitigations: require explicit user approval before running tests, run test execution only in an isolated sandbox with no access to host credentials and restricted network egress, prefer verifying test output locally rather than collecting environment-sensitive logs, and enforce use of pinned tool versions / lockfiles before using npx. With these mitigations, the skill can be used safely for intended test-generation tasks.