The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill presents an indirect prompt injection surface as it ingests and processes data from external API responses while possessing write capabilities like asset transfers and account freezing. However, this is adequately mitigated by the developer's security instructions. Ingestion points include the fetchSigned function in SKILL.md which parses API response data. Boundary markers are present in the form of mandatory human 'CONFIRM' prompts for all write operations on live accounts. Capability inventory includes network operations via fetch for financial transfers and API key management. Sanitization is performed by the validateParams function, which explicitly filters for injection-related characters such as ampersands, equals signs, and newlines.
[EXTERNAL_DOWNLOADS]: The skill identifies json-bigint as a necessary dependency for correctly parsing large integer values returned by the BingX API, which is a standard and safe practice for financial applications.
[COMMAND_EXECUTION]: No dangerous shell command execution or privilege escalation patterns were detected in the skill instructions or helper code.

bingx-sub-account