The Agent Skills Directory

[COMMAND_EXECUTION]: The skill suggests using shell commands for diagnostic instrumentation during the root cause investigation phase. It provides examples such as env | grep IDENTITY and macOS security commands to help the agent identify component failures. These are legitimate debugging tools used within the agent's execution context.
[SAFE]: No evidence of data exfiltration, obfuscation, or remote code execution was found. The instructions are aligned with the skill's stated purpose of improving technical troubleshooting.
[PROMPT_INJECTION]: The skill uses authoritative language (e.g., "The Iron Law", "MUST complete") to define its debugging process. This is instructional rather than an attempt to bypass the underlying model's safety guardrails or system instructions.

systematic-debugging