xy-idea-desk
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill facilitates data retrieval by executing a local Python script located at
scripts/atoms-search.py. This script is used to search for 'atoms' (knowledge snippets) to provide evidence for topic evaluation. - [EXTERNAL_DOWNLOADS]: The search script is configured to perform network requests to
https://api.xyskill.xyz. This domain belongs to the vendor's infrastructure and is used as a cloud-based source for the knowledge repository. The script includes a fail-safe to revert to local file access if the cloud service is unreachable. - [INDIRECT_PROMPT_INJECTION]: The skill handles untrusted data from both user prompts and external search results. To mitigate risks, the instructions explicitly prohibit the agent from hallucinating results ('绝不现场编一个听起来合理的数字') and require clear attribution to the source database ('依据：原子 id {本轮引用}').
- [DATA_EXPOSURE]: The search script attempts to read a configuration file at
~/.xy/config.json. This is used to locate the installation root for localized resources and represents vendor-specific configuration rather than the exposure of sensitive system credentials.
Audit Metadata