douyin-account-diagnosis
Pass
Audited by Gen Agent Trust Hub on Oct 28, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches account and video data from the vendor's API at https://redfox.hk/story/api/dyUser/queryData. This is a core part of the skill's functionality and utilizes the author's own infrastructure. These requests are documented as safe vendor-resource interactions.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Douyin, including account signatures and video descriptions. This content is interpolated into the agent's context in the form of a diagnostic report. There are no explicit boundary markers or instructions to the agent to ignore any embedded commands within this third-party data, creating a surface for indirect prompt injection.
- Ingestion points: Data is retrieved from the Douyin API via scripts/douyin_diagnosis.py and scripts/douyin_api_client.py.
- Boundary markers: Absent. Data is presented in standard Markdown tables and lists without specific delimiters or warnings.
- Capability inventory: The skill performs network requests (requests and urllib.request) and data processing but does not execute arbitrary system commands, perform local file writes, or access sensitive system credentials.
- Sanitization: Content is truncated (e.g., descriptions are cut to 60 characters), but no specific sanitization or escaping for malicious prompt patterns is performed.
Audit Metadata