The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple shell and PowerShell commands to perform builds, tests, and environment setup.
Executes dotnet build, dotnet test, and repository-specific scripts like .\build.ps1 and .\scripts\setup-dev.ps1.
Uses the crabbox CLI (a tool from the vendor openclaw.ai) to run commands on remote hosts.
[PRIVILEGE_ESCALATION]: The skill instructs the agent to bypass PowerShell execution policies.
Uses powershell.exe -ExecutionPolicy Bypass to launch applications locally on remote hosts as specified in SKILL.md.
[DYNAMIC_EXECUTION]: The skill generates and executes PowerShell and Bash scripts dynamically at runtime.
Employs here-doc patterns (<<'POWERSHELL', <<'BASH') to pass multi-line scripts to the crabbox run --script-stdin command in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from the current repository checkout and executes it on remote infrastructure, creating a vector for instructions embedded in code to influence agent behavior.
Ingestion points: Files within the current git checkout (SKILL.md).
Boundary markers: Absent.
Capability inventory: Execution of arbitrary PowerShell/Bash scripts via crabbox run and local shells (SKILL.md).
Sanitization: Absent; while the skill warns users to review untrusted code before syncing, there is no automated sanitization of repository content.

crabbox