The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The installation instructions fetch a shell script from a remote URL and pipe it directly into the shell for execution.
Evidence: curl -fsSL https://raw.githubusercontent.com/okx/onchainos-skills/main/install.sh | sh in SKILL.md.
[EXTERNAL_DOWNLOADS]: The skill downloads a pre-compiled binary (gmx-v2) and other CLI tools from the vendor's GitHub repository during the setup process.
Evidence: curl -fsSL "https://github.com/okx/plugin-store/releases/download/plugins/gmx-v2@0.2.1/gmx-v2-${TARGET}${EXT}" in SKILL.md.
[DATA_EXFILTRATION]: The installation script performs device fingerprinting by collecting the system hostname, operating system, architecture, and user's home directory path. This data is sent to external telemetry and vendor endpoints.
Evidence: DEV_RAW="$(hostname):$(uname -s):$(uname -m):$HOME" used in POST requests to https://plugin-store-dun.vercel.app/install and https://www.okx.com/priapi/v1/wallet/plugins/download/report.
The script uses Base64 encoding to mask an internal HMAC signature key used for reporting.
[COMMAND_EXECUTION]: The skill's implementation relies on executing shell commands via the onchainos platform CLI and the project-specific gmx-v2 binary to perform blockchain operations.
[PROMPT_INJECTION]: The skill retrieves and displays data from external blockchain RPC providers and GMX APIs. This untrusted content serves as a potential surface for indirect prompt injection attacks.
Mitigation: The developer included a security notice regarding untrusted external content in SKILL.md.

gmx-v2