The Agent Skills Directory

[COMMAND_EXECUTION]: The tutorial explicitly instructs users to set tools.profile to full to fix "AI can't do anything" issues. This setting enables a complete toolset that includes shell command execution capabilities for the agent.
[INDIRECT_PROMPT_INJECTION]: The skill documents an architecture where the AI assistant ingests untrusted data from various external sources, which could contain malicious instructions.
Ingestion points: The skill describes tools for reading web pages (openclaw infer web), processing incoming chat messages via webhooks from platforms like Feishu, Slack, and WeChat, and ingesting documents into a knowledge base (openclaw knowledge add).
Boundary markers: No explicit boundary markers or "ignore embedded instructions" warnings are documented in the configuration examples or integration code.
Capability inventory: When configured as recommended (tools.profile full), the agent has the capability to execute shell commands, perform file management, and initiate network requests.
Sanitization: The provided JavaScript integration examples (e.g., the Feishu webhook handler) pass raw user content directly to the OpenClaw chat API without sanitization or validation.
[EXTERNAL_DOWNLOADS]: The installation instructions direct users to download and install the openclaw package and its SDK from public registries using npm, yarn, and docker. While these are standard package managers, the safety of the skill relies on the integrity of the third-party openclaw repository and its maintainers.

awesome-openclaw-tutorial