scorched-earth

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill mandates exhaustive extraction of "every specific" detail and exact file/line findings (and to write those findings to disk) with no guidance to redact sensitive values, so any embedded API keys, tokens, or passwords discovered would be expected to be captured and output verbatim — creating a high exfiltration risk.

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

• Third-party content exposure detected (high risk: 0.85). This skill’s Intake Mode (/scorched-earth [url]) and Research Mode (/scorched-earth research [topic]) explicitly fetch and read outsider-authored web content / docs / repos at runtime (e.g., arbitrary URLs, public pages, GitHub repos), which the LLM then ingests as readable text for extraction.