The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The setup.sh script attempts to use the sudo command to install the gate-cli binary into /usr/local/bin if it cannot write to the user's local directory. This requests elevated permissions during the setup process.
[EXTERNAL_DOWNLOADS]: The skill downloads the gate-cli binary directly from the official Gate GitHub repository (github.com/gate/gate-cli) to provide core functionality. This is a standard vendor-provided dependency for the skill's operation.
[COMMAND_EXECUTION]: The skill relies on executing the gate-cli binary to perform all sub-account operations, including listing, creating, and locking accounts. Commands are executed with parameters derived from user input.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data retrieved from external API calls, specifically sub-account login names and remarks, and displays them to the agent. This presents an attack surface where an attacker could embed malicious instructions within sub-account metadata to influence the agent's behavior.
Ingestion points: API responses from gate-cli cex sub-account get and list subcommands (referenced in SKILL.md).
Boundary markers: No specific delimiters or "ignore instructions" warnings are applied to the displayed sub-account metadata in the output templates.
Capability inventory: The skill possesses command execution capabilities (gate-cli) and is authorized to perform state-changing operations (create/lock/unlock) on sub-accounts.
Sanitization: No explicit sanitization or filtering of account remarks or login names is documented before the data is presented in the agent's context.

gate-exchange-subaccount