zhihe-legal-research
Warn
Audited by Gen Agent Trust Hub on Apr 23, 2026
Risk Level: MEDIUMDATA_EXFILTRATIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill collects user phone numbers and legal queries, transmitting them to a remote API at
fc-openresearch-qzquocekez.cn-shanghai.fcapp.runfor authentication and research purposes. - [DATA_EXFILTRATION]: Multiple scripts access and manipulate sensitive files outside the skill's own directory, including
~/.zhihe-legal-research/configand~/.openclaw/.env, to read or migrate authentication tokens. - [COMMAND_EXECUTION]: The skill uses
python3 -cto dynamically process user-provided strings and generate filenames at runtime withinscripts/research.shandscripts/monitor.sh. - [COMMAND_EXECUTION]: Shell scripts utilize
sed -ito programmatically modify configuration files containing authentication tokens. - [EXTERNAL_DOWNLOADS]: The skill downloads legal research reports (docx files) from external URLs provided by the remote API during the archiving process.
- [PROMPT_INJECTION]: The skill processes and displays research results from an external API, creating a surface for indirect prompt injection.
- Ingestion points:
scripts/research.shandscripts/monitor.shingesttext_resultand report content from API responses. - Boundary markers: None observed in the processing of external API data.
- Capability inventory: Subprocess execution (
curl,python3,sed), file system operations (mkdir,rm,chmod), and network access viacurl. - Sanitization: Python scripts are used to sanitize filenames, but no sanitization is applied to the textual research results displayed to the user.
Audit Metadata