Please report vulnerabilities privately through GitHub security advisories, not in a public issue. Include what you found, how to reproduce it, and the impact you expect. You will get a response within a week, and a fix or a plan for one as soon as the issue is understood.

Security fixes go into the latest release.

ThinkLess sits between user input and actions your agent takes, so the following are in scope:

• A way for input to change which provider answers a question, to inject an answer, or to bypass a question's allowed answers.
• Trace content leaking when capture_content is off, or the OpenTelemetry sink exporting content it should drop.
• Credentials appearing in traces, logs or error messages.
• Unsafe deserialization or file handling in trace, pricing or dataset loading.

Model quality issues, such as a decision model answering a question wrongly, are not vulnerabilities; please open a regular issue with the example. Keep consequential actions behind deterministic checks, as described in the production guide.