The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The setup.sh installation script attempts to acquire administrative privileges by using sudo install to place the gate-cli binary into /usr/local/bin if user-level directory installation fails. This allows the skill to perform operations with root-level access on the host system during setup.
[EXTERNAL_DOWNLOADS]: The skill fetches the gate-cli binary from the official vendor repository at github.com/gate/gate-cli during the installation process defined in setup.sh and the SKILL.md frontmatter. This is part of the required toolchain setup for the skill to function.
[COMMAND_EXECUTION]: The skill executes the gate-cli binary using shell commands to perform financial operations on the Gate exchange. The resolution of this binary follows a specific search order on the host filesystem as defined in SKILL.md and gate-runtime-rules.md.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data returned from external API responses via the gate-cli tool.
Ingestion points: Data enters the agent's context through command outputs from tools like get_unified_accounts and list_unified_loan_records.
Boundary markers: A mandatory confirmation gate is enforced; the agent must present an "Action Draft" and receive explicit user confirmation before executing any mutation command.
Capability inventory: The skill has capabilities to execute shell commands (gate-cli) and perform account modifications (borrowing, repayment, and configuration changes).
Sanitization: The skill instructions mandate exact preservation of API numeric strings and manual review via human-in-the-loop confirmation steps for all high-risk operations.

gate-exchange-unified