The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/reduce_test.py script enables the execution of arbitrary system commands.
Evidence: The TestCaseReducer.run_test method in scripts/reduce_test.py uses subprocess.run to execute commands passed via the --command parameter. This allows the agent to execute any available system command with arbitrary arguments.
[PROMPT_INJECTION]: The skill exhibits a vulnerability to indirect prompt injection when processing and executing content from external, potentially untrusted files.
Ingestion points: The scripts/reduce_test.py script reads and writes the contents of the file specified by the test_file positional argument during the reduction process.
Boundary markers: There are no boundary markers, delimiters, or explicit instructions within the skill to prevent the agent from obeying instructions that might be embedded in the test files being processed.
Capability inventory: The skill possesses the capability to write to the local filesystem and execute arbitrary shell commands via the subprocess module.
Sanitization: No sanitization, validation, or sandboxing is applied to the test file content or the execution command, meaning any malicious code contained within a test case will be executed with the same privileges as the agent during the reduction process.

test-case-reducer