gracker-deep-research

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly instructs scanning local files with rg and embedding the extracted file snippets verbatim into local-context, references, and master-research files (and then using those files for writing), which will cause any secrets present in those files to be included in LLM inputs/outputs and thus can be exfiltrated.

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

• Third-party content exposure detected (high risk: 0.78). Phase 1c “参考资料归档”会把研究员抓取/打开得到的网页全文（{full content}）写入本地 references/*.md，而 Phase 2b/2c 随后会读取这些 references/*.md 作为写作/质检上下文输入（间接承载外部自由文本）。