The Agent Skills Directory

[SAFE]: The skill implements secure local storage for sensitive Veil keys, using environment files with restricted permissions (chmod 600) to protect against unauthorized access.
[COMMAND_EXECUTION]: The skill wraps the @veil-cash/sdk CLI tool to execute shielded transactions. This is the primary intended function of the skill and is performed using local script execution.
[EXTERNAL_DOWNLOADS]: All external dependencies are standard tools or libraries. The user is instructed to install them from official repositories, and the skill does not perform any unverified background downloads.
[DATA_EXFILTRATION]: Outbound network traffic is directed to the user's RPC provider and the vendor's Bankr API (api.bankr.bot). These communication channels are necessary for the skill's functionality and are documented neutrally.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface where user-provided input is interpolated into requests sent to the Bankr Agent API.
Ingestion points: Arguments passed to scripts/veil-bankr-prompt.sh and transaction JSON in scripts/veil-bankr-submit-tx.sh.
Boundary markers: Some scripts use a prefix to instruct the recipient API not to change fields, but most forward raw input.
Capability inventory: Network access via curl and local command execution via the Veil CLI.
Sanitization: Data is JSON-escaped for transport, but the skill does not sanitize content for adversarial instructions directed at the downstream API.

veil