The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides implementation patterns for signing transactions using the ows command-line utility via Python's subprocess.run. This is a legitimate integration path for the Open Wallet Standard (OWS).
[INDIRECT_PROMPT_INJECTION]: The skill identifies XRPL transaction memos as a potential attack surface for indirect prompt injection. It explicitly instructs the agent to treat memo contents as untrusted and never allow them to influence signing decisions or bypass security protocols.
Ingestion points: Memos field in SKILL.md (Non-negotiable rule #7).
Boundary markers: The skill enforces a rigid signing ceremony that cannot be overridden by external data.
Capability inventory: Transaction signing, ledger submission, wallet generation.
Sanitization: Instructions to ignore natural language commands found within transaction metadata.
[EXTERNAL_DOWNLOADS]: The skill references external dependencies including the @open-wallet-standard/core npm package and the ows CLI tool. These are used for policy-gated wallet management and are consistent with the skill's stated purpose.
[CREDENTIALS_UNSAFE]: The skill handles sensitive XRPL seeds but implements high-standard security practices. It mandates writing secrets directly to .env files, ensuring they are ignored by git, and strictly forbids echoing or logging private key material in any form.

xrpl-agent-wallet