The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of arbitrary scripts within a managed environment through the --setup-script parameter in environment creation commands (e.g., arkcli agent env create --setup-script @./bootstrap.sh). While this is a feature of the managed service, it provides a direct path for the agent to execute code.
[EXTERNAL_DOWNLOADS]: The skill provides functionality to scan and import agent skills directly from GitHub repositories (e.g., arkcli agent skill github import owner/repo), which allows for the introduction of external code into the agent's environment from arbitrary sources.
[DATA_EXFILTRATION]: The skill includes multiple commands for uploading local files to a remote platform (arkcli agent file upload) and mounting them into sessions (arkcli agent session resources add --path <file>). The @file syntax enables the agent to read local files, which creates a data exposure risk if the agent is directed to access sensitive files (e.g., SSH keys or environment variables).
[CREDENTIALS_UNSAFE]: The skill provides extensive capabilities for managing sensitive credentials, including API keys, OAuth tokens, and Vault secrets. These commands (e.g., arkcli agent vault credentials create) handle high-value secrets that could be targeted for exposure if the agent's instructions are compromised.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a broad attack surface for indirect prompt injection as it ingests and processes data from external sources such as the Skill Market, GitHub repositories, and session event streams.

arkcli-agent