The Agent Skills Directory

[PERSISTENCE]: The skill provides instructions for establishing a persistent background service on macOS using openclaw gateway install. This command sets up a per-user LaunchAgent, which is the intended mechanism to ensure the local agent remains active across system sessions.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the setup of an AI agent host (OpenClaw) that ingests external data and prompts, creating a surface for indirect prompt injection attacks.
Ingestion points: The agent is designed to process user prompts and agent workflows through its gateway and dashboard interfaces (SKILL.md).
Boundary markers: The documentation recommends using a 'synthetic prompt' to test tool permissions and provider access (SKILL.md), which serves as a basic control.
Capability inventory: The setup grants the agent access to manage the gateway service and perform host inventory tasks, such as reading power settings (pmset) and software versions (references/details.md).
Sanitization: The documentation does not specify sanitization or output filtering methods for the processed prompts.
[COMMAND_EXECUTION]: The skill lists several commands for the agent to execute, including openclaw gateway install for service management and pmset -g for host inventory. The documentation provides these within a security-conscious context, advising that all actions be authorized and reviewed prior to execution.

openclaw-local-mac-mini