The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to perform a global installation of the openclaw package from the NPM registry (npm install -g openclaw@latest).
[COMMAND_EXECUTION]: The skill executes multiple CLI commands to manage the gateway and its environment.
openclaw onboard --install-daemon: Configures a system background service, establishing persistence and potentially requiring elevated privileges.
openclaw cron add: Sets up recurring automated tasks that persist on the host system.
[DATA_EXFILTRATION]: The gateway is designed to communicate with external messaging platforms and webhooks.
Configures integrations with WhatsApp, Telegram, and Discord for data transmission.
Exposes webhook endpoints to receive triggers from external sources like CI/CD pipelines.
[PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by processing untrusted data from external communication channels.
Ingestion points: Incoming messages from WhatsApp, Telegram, Discord, and webhook payloads.
Boundary markers: The skill lacks explicit instructions or delimiters to isolate untrusted message content from the agent's instructions.
Capability inventory: Access to the host's CLI for command execution and file system access within the ~/.openclaw/ directory.
Sanitization: There are no mechanisms defined to sanitize or validate external content before it is processed by the AI agents.

openclaw