The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use the exec tool for running shell commands during data discovery and application execution. This includes examples for system monitoring and process management.
[REMOTE_CODE_EXECUTION]: The established workflow involves writing Node.js scripts to the local filesystem and subsequently executing them using the shell. This pattern enables the generation and execution of arbitrary code.
[CREDENTIALS_UNSAFE]: The skill explicitly directs the agent to interact with sensitive environment files, specifically instructing it to read and verify secrets from ~/.openclaw/workspace/.env.
[DATA_EXFILTRATION]: The skill encourages creating scripts that interface with external APIs to send or receive data, providing a vector for transferring local information to remote services.
[PROMPT_INJECTION]: The skill processes external data from shell outputs and database records which could contain malicious instructions.
Ingestion points: Data returned from Query("exec"), Query("read"), and Query("db_query").
Boundary markers: No specific delimiters or "ignore instructions" warnings are defined for the DSL runtime.
Capability inventory: exec, write, db_execute, and app_update capabilities are available across the skill's toolset.
Sanitization: The instructions provide a security best practice of using database parameters to avoid string-based shell injection, though no platform-level enforcement is mentioned.

openui-app