audit-book-writing-v1

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

• Third-party content exposure detected (high risk: 0.72). The skill’s runtime workflow includes “Material Collection” steps that download/search “Practical Cases” and “Regulatory Documents” from external websites (e.g., regulator/enforcement pages) and then uses their extracted text as LLM context for writing case/regulatory chapters, which is outsider-authored free text.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill includes explicit runtime fetch commands (e.g., WebFetch "https://www.ncha.gov.cn/art/2023/...") and WebSearch/WebFetch workflows to retrieve authoritative webpages for data verification that are then injected into the agent's context to drive citations and writing, so external content from https://www.ncha.gov.cn/art/2023/... is used at runtime to directly influence prompts/output.

MEDIUM W021: Hidden or invisible Unicode characters detected (potential obfuscation or prompt injection).

• Hidden Unicode characters detected (1 type(s) found)