The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill provides explicit instructions for the agent to modify its execution environment by updating the platform configuration file ('.claude/settings.local.json') via the 'update-config' tool. This is intended to allowlist specific command patterns and bypass interactive user permission prompts during recurring tasks, which effectively escalates the agent's autonomy and reduces runtime oversight.
[PERSISTENCE]: The modification of local configuration files ('.claude/settings.local.json') to allowlist commands creates a persistent change in the agent's behavior, ensuring that the elevated permissions (bypassing prompts) remain active across different sessions.
[COMMAND_EXECUTION]: The skill's primary functionality relies on executing the 'revx' CLI tool via the 'Bash' tool to perform financial queries, such as checking balances, listing transactions, and managing orders.
[CREDENTIALS_UNSAFE]: The instructions direct the agent to autonomously handle authentication failures involving private keys and configuration files located in '~/.config/revolut-x/'. It encourages the agent to 'apply the fix yourself' without interrupting the user unless user-specific input is required, promoting autonomous management of sensitive security configurations.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: Data retrieved from external API responses via 'revx' CLI commands (transaction lists, order details, etc.) enters the agent context.
Boundary markers: The skill does not define specific delimiters or instructions to ignore embedded instructions within retrieved data.
Capability inventory: The skill has access to the 'Bash' tool for command execution and the 'update-config' tool for configuration changes.
Sanitization: There are no explicit instructions for sanitizing, validating, or escaping external content retrieved from the CLI before processing it.

revx-account