expert-xhs
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
subprocess.runinxhs_engagement.pyto invoke thecamoufox-clibrowser automation tool and an internal metrics update script. The commands are constructed using static strings and validated session names, minimizing injection risks. This execution is fundamental to the skill's purpose of scraping engagement data from the creator backend. - [INDIRECT_PROMPT_INJECTION]: The skill processes external data by downloading and analyzing content from Xiaohongshu notes via
xhs-content-opsandxhs-style-profiler. - Ingestion points: External note text, titles, and engagement statistics are ingested during the benchmark and profiling workflows.
- Boundary markers: The instructions include explicit rules for data integrity, requiring the agent to verify signals against raw text and prohibiting the fabrication of data.
- Capability inventory: The skill has the capability to perform network operations, write to the workspace (DNA and output directories), and execute shell commands for publishing and scraping.
- Sanitization: The Python analysis scripts use regular expressions and normalization techniques to extract structured data from raw content.
- [OBFUSCATION]: The script
publish_xhs.pyutilizes the Unicode character\uFEFF(Byte Order Mark). This is a functional requirement of the Xiaohongshu web API to correctly identify inline hashtags and is not used for malicious concealment.
Audit Metadata