The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documents the openshell doctor exec command, which allows users to execute arbitrary shell commands inside the gateway container environment for diagnostic and administrative purposes.
[DATA_EXFILTRATION]: The instructions reference the use of local sensitive files, such as SSH private keys (~/.ssh/id_rsa), which are required for the standard workflow of provisioning and managing remote gateways.
[PROMPT_INJECTION]: The skill describes a process for monitoring sandbox logs (openshell logs) to observe denied network actions and refine policies. This introduces a surface where untrusted data from the sandbox environment is ingested into the agent's context.
Ingestion points: Untrusted log content is read via the openshell logs command as part of the policy iteration workflow in SKILL.md.
Boundary markers: The documentation does not implement specific boundary markers or instructions to ignore potential commands embedded in logs.
Capability inventory: The agent has access to CLI tools for infrastructure management, command execution, and policy updates.
Sanitization: There is no description of sanitizing or validating log output before processing.

openshell-cli