rtfp

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt requires emitting verbatim transcript fields (assistant_excerpt and user_reply) into generated outputs/files, which forces the model to reproduce any secrets present in the transcripts verbatim and thus risks exfiltration.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.65). Outsider-authored free text can enter the LLM context during Stage 3 “Context Reconstruction,” because the frustration-analyzer:context-reconstructor reads the full transcript context for the selected winner from the chosen session file (which is not guaranteed to be user-authored), then reconstructs and excerpts the triggering assistant output and user reply for the LLM to process.