The Agent Skills Directory

[COMMAND_EXECUTION]: The gradle_run.py script executes shell commands using subprocess.Popen to perform Gradle build and test tasks. This is the primary purpose of the skill. The script includes several safety mechanisms, such as verifying the command starts with a recognized Gradle launcher and executing the tasks in isolated process groups.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data by reading and summarizing Gradle build output, which represents a potential attack surface for indirect prompt injection if the build logs contain malicious instructions intended for the agent.
Ingestion points: The script reads build output from log files stored in a managed subdirectory within the OS temporary directory.
Boundary markers: Instead of raw output, the tool provides the agent with structured JSON summaries and bounded excerpts.
Capability inventory: The skill facilitates execution of local Gradle commands, writing of diagnostic ledgers, and cleanup of its own temporary log files.
Sanitization: The script includes a robust redaction mechanism that uses regular expressions to mask common sensitive patterns, such as authorization headers, passwords, and API tokens, before the data reaches the agent context.

gradle-run