The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install several specialized packages from the NPM registry, including @glamsystems/glam-cli, @glamsystems/glam-sdk, and @glamsystems/ix-mapper, to enable vault management and DeFi interactions.
[COMMAND_EXECUTION]: The skill utilizes the glam-cli command-line tool to execute a wide variety of Solana vault operations, ranging from administrative configuration to complex DeFi swaps and lending.
[DATA_EXFILTRATION]: The skill manages sensitive Solana keypair files (e.g., ~/.config/solana/id.json) and instructs users to store secrets like API keys in local configuration files, creating a potential point of exposure for cryptographic credentials.
[DYNAMIC_EXECUTION]: The skill uses the ix-mapper library to dynamically transform and remap Solana transaction instructions at runtime, which is a core component of the protocol's access control architecture.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface through the ingestion of external data. 1. Ingestion points: The skill parses user-provided JSON templates for vault creation and retrieves live data from Solana RPC nodes and the Jupiter API. 2. Boundary markers: There are no explicit delimiters or warnings defined to prevent the agent from executing instructions embedded in this external data. 3. Capability inventory: The skill can execute shell commands via glam-cli and perform network operations using the Solana and Jupiter APIs. 4. Sanitization: The instructions do not detail any specific input validation or sanitization steps, relying on the internal logic of the protocol's packages.

glam