collusionscreen-collusion-screening-free
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill documentation provides instructions for the agent to install the
@alipay/agent-paymentpackage vianpx. This command fetches and executes code from the official Alipay organization on the npm registry to facilitate the installation of an upgraded version of the skill. This is a standard integration for the SkillPay platform and utilizes a well-known service provider.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes text data from multiple bidders to identify overlaps in contact information and project staff, creating a potential surface for indirect prompt injection.\n - Ingestion points: The skill reads untrusted text from the
textfield of the input JSON, which is processed by the analysis engine inscripts/engine/collusion-screening.js.\n - Boundary markers: The skill does not employ delimiters or specific instructions to isolate untrusted bidder text from the internal analysis context.\n
- Capability inventory: The skill environment has access to
ReadandBashtools. The Node.js engine script is self-contained and lacks capabilities for network communication or file system modification.\n - Sanitization: The analysis engine uses rigid regular expression patterns to extract data (such as phone numbers and names). This ensures that the extracted strings are treated as data values rather than instructions, mitigating the risk of exploitation.
Audit Metadata