secret-in-result: agent-bound text (tool result, resource, prompt, …) matching a known secret format → secret-exposure (high). Evidence is redacted before recording. Covers every protocol surface, not just tools/call.
secret-in-result
secret-in-result: agent-bound text (tool result, resource, prompt, …) matching a known secret format → secret-exposure (high). Evidence is redacted before recording. Covers every protocol surface, not just tools/call.