The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for executing shell commands on remote AWS, Blacksmith Testbox, and local Docker environments using the crabbox run command.
[REMOTE_CODE_EXECUTION]: Local scripts and shell instructions are uploaded and executed on remote target hosts to perform end-to-end testing and validation.
[EXTERNAL_DOWNLOADS]: The skill references installing the crabbox CLI from the author's Homebrew tap (openclaw/tap/crabbox) and utilizes remote coordination services at crabbox.openclaw.ai.
[PROMPT_INJECTION]: An indirect prompt injection surface exists when the tool fetches and executes code from GitHub Pull Requests using the --fresh-pr flag.
Ingestion points: The skill ingests untrusted code from GitHub PRs via the crabbox run --fresh-pr command.
Boundary markers: There are no explicit instructions to the agent to treat PR content as untrusted data or to ignore embedded instructions.
Capability inventory: The environment allows for remote shell access, script execution, and desktop UI manipulation via WebVNC.
Sanitization: No explicit sanitization or pre-execution validation of PR content is described in the workflow.

crabbox