The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads the gate-cli binary from the official vendor's release pages on GitHub. This is the legitimate installation process for the required toolset.
Evidence: SKILL.md frontmatter (install section) and setup.sh download logic.
[COMMAND_EXECUTION]: The skill utilizes 58 distinct subcommands of the gate-cli tool to query account status and manage assets. All operations that modify account state (write operations) are gated by a mandatory 'Strong Confirmation' mechanism requiring human approval.
Evidence: SKILL.md (gate-cli commands used) and references/scenarios.md (Global Execution Gate).
[PRIVILEGE_ESCALATION]: The setup.sh installation script includes logic to use sudo install as a fallback if the standard installation to the user's local directory fails. This is a common pattern for installing CLI tools to system paths.
Evidence: setup.sh (install_bin function).
[INDIRECT_PROMPT_INJECTION]: As an orchestration skill processing API data and user inputs, there is an inherent vulnerability surface for indirect prompt injection. This risk is effectively mitigated by the skill's architecture, which requires explicit user review of an 'Action Draft' before any execution occurs.
Ingestion points: SKILL.md (API results from gate-cli) and references/scenarios.md (user-supplied parameters).
Boundary markers: Data is processed into structured templates, though raw API results do not use explicit delimiters.
Capability inventory: 54 read tools and 4 write tools associated with account and asset management.
Sanitization: Parameter validation is handled by the underlying gate-cli binary.

gate-exchange-assets-manager