The Agent Skills Directory

Automated Tool Approvals: The skill configures environment variables such as SHELL_AUTO_APPROVE and AUTO_APPROVE_MCP to enable the agent to execute commands without manual intervention. This is a functional requirement for unattended automation but reduces the oversight typically provided by human-in-the-loop checkpoints.
Command Execution from Local State: The rerun functionality determines which scripts to execute based on entries in a local temporary file (.tmp/examples-rerun.txt). This relies on the integrity of the data collected from logs during previous runs to construct shell commands.
Indirect Prompt Injection Surface: The skill directs the agent to analyze external source code and log files to validate execution results. Reading unstructured data from these files creates a surface where the agent might interpret content within logs as instructions during the validation phase.

examples-auto-run