The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains phrases like 'ignore previous instructions' and 'override detection level', but these are used in a purely defensive context within the 'Prompt Injection Safety Notice' section. This section instructs the agent to ignore such commands if they appear in user-provided logs or reports, which is a recommended hardening practice.
[DATA_EXFILTRATION]: The skill instructions proactively advise the agent to redact sensitive information such as internal IP addresses and usernames from its output. No network-capable tools or exfiltration patterns were detected.
[COMMAND_EXECUTION]: Access is restricted to basic file tools (Read, Grep, Glob) via the 'allowed-tools' field. The skill explicitly forbids the agent from executing any scripts or commands found within the technical content it analyzes.

detection-engineering