The Agent Skills Directory

[COMMAND_EXECUTION]: Several scripts, such as agenthub/scripts/session_manager.py, autoresearch-agent/scripts/run_experiment.py, and skill-tester/scripts/script_tester.py, utilize the subprocess module to run local commands. These operations include interacting with git for repository management, executing script-based tests, and running user-defined benchmarking commands. These actions are consistent with the tools' stated purposes of automation and performance optimization.
[PROMPT_INJECTION]: Documentation files in the skill-security-auditor module (e.g., references/threat-model.md) contain examples of prompt injection strings like 'Ignore previous instructions'. These are documented as threat patterns for the tool to detect and do not represent active attempts to bypass safety filters in the current context.
[SAFE]: The skill primarily uses the Python standard library and does not include obfuscated functional code or unauthorized network communication. No hardcoded credentials or data exfiltration patterns were found. The skill operates on local project data to provide engineering insights.

engineering-advanced-skills