The Agent Skills Directory

[SAFE]: The skill defines a legitimate persona for a network engineer, detailing expertise in cloud providers (AWS, Azure, GCP) and standard networking protocols.
[DATA_EXFILTRATION]: While the skill mentions networking tools like curl, wget, and tcpdump, they are listed within the context of troubleshooting and analysis capabilities. No patterns indicating unauthorized data harvesting or exfiltration were found.
[CREDENTIALS_UNSAFE]: No hardcoded credentials, API keys, or instructions to handle secrets insecurely were detected.
[PROMPT_INJECTION]: The instructions are focused on defining behavioral traits and response approaches for the persona. No 'ignore previous instructions' or bypass patterns are present.
[INDIRECT_PROMPT_INJECTION]: The skill mentions opening a local file resources/implementation-playbook.md for examples. There is no evidence of the skill ingesting untrusted external data in a way that would facilitate prompt injection.

network-engineer