The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides shell commands involving ssh, pnpm, and prlctl. It instructs the agent to use ssh to execute commands on a specific remote host (peters-mac-studio-1) and utilizes prlctl exec to run shell commands inside guest VM environments with elevated permissions.
[CREDENTIALS_UNSAFE]: The skill includes logic to retrieve and use authentication tokens. It provides a shell command that extracts a Discord token from a configuration file located at ~/.openclaw/openclaw.json on a remote machine and references an OPENAI_API_KEY as an environment input.
[DATA_EXFILTRATION]: The skill retrieves authentication tokens from a hidden configuration file on a remote server. The instruction to ssh into a host and run jq to read a token from ~/.openclaw/openclaw.json describes a mechanism for accessing secrets on a remote file system.
[EXTERNAL_DOWNLOADS]: The skill uses pnpm, which typically involves downloading external Node.js packages during the execution of test suites.
[SAFE]: The skill presents an indirect prompt injection surface as it reads data from Discord channel history. 1. Ingestion points: Reads Discord channel history using the openclaw message read command. 2. Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore embedded commands in the channel data. 3. Capability inventory: The skill utilizes subprocess calls for VM control (prlctl), remote shell access (ssh), and package management (pnpm). 4. Sanitization: No sanitization or validation logic for the external Discord content is specified in the skill file.

parallels-discord-roundtrip