brain

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill explicitly requires copying short exact quotes from captures (using get-capture with includeRawContent when available), which forces verbatim reproduction of stored text that could contain secrets, even though transcript sanitization will strip API-key-shaped strings by default.

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

• Third-party content exposure detected (high risk: 0.75). SKILL.md describes runtime ingestion of meeting-shaped transcripts via import-transcript/import-capture (with sourceUrl and transcript text) and later distillation using get-capture/write-knowledge, which can include outsider-authored free text from external sources the operating user did not author.