experiment-journal

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.95). The prompt explicitly instructs the agent to read and embed raw files (e.g., results-summary.json, run-config.json, full logs and transcripts) verbatim into the journal (including an appendix with the full results-summary.json), which would force the LLM to output any secrets present in those files.