testskill

This skill appears internally consistent and lightweight, but be cautious when running tests: 1) Do not paste secrets, private keys, passwords, or confidential documents into test inputs. 2) Avoid submitting very large datasets without understanding cost/latency implications. 3) Review agent outputs before acting on them—benchmarks or adversarial inputs may provoke unexpected or incorrect responses. 4) If you want stricter limits, disable autonomous invocation or run tests in isolated conversations where the agent can't access other context outside the test. If you need more certainty, ask the author for example test payloads and explicit data handling / retention policies.