The Agent Skills Directory

[COMMAND_EXECUTION]: The instructions in 'references/gradle.md' describe building shell commands using parameters like test filters and module names. Without explicit sanitization of these inputs, there is a risk of command injection if malicious strings are provided as input.
[DATA_EXFILTRATION]: The script 'scripts/run_tests_in_ide.kts' uses an unconfigured XML parser to read test results. This is vulnerable to XML External Entity (XXE) attacks, which could allow an attacker to read files from the local file system if they can influence the contents of the 'build/test-results/' directory.
[PROMPT_INJECTION]: The skill exhibits an Indirect Prompt Injection surface by ingesting untrusted data from JUnit XML reports. Ingestion points: JUnit XML files in 'build/test-results/'. Boundary markers: Absent. Capability inventory: Shell command execution and IDE script execution. Sanitization: Absent. This allows for potential instructions embedded in test failures to influence the agent.

run-tests