lore

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt requires the agent to display verbatim "evidence" and real command sequences from user transcripts (dossiers and plan artifacts) in chat, which can include API keys, tokens, or passwords; although a later privacy-scrub step is suggested for SKILL.md files, the required prior verbatim dossier rendering forces the LLM to handle/output secrets directly.

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

• Third-party content exposure detected (high risk: 0.85). At runtime, the deep-mine workflow (scripts/deep-mine.workflow.js) runs node ... beats ... to export beat spans, including text (free-form transcript content) from the user’s SpecStory history files, and passes that readable text into the LLM agents for dossier generation.