The Agent Skills Directory

[PROMPT_INJECTION]: No evidence of prompt injection or instruction overrides. The text defines a clear professional persona without attempting to bypass safety filters.
[DATA_EXFILTRATION]: No sensitive file paths, credential markers, or suspicious network exfiltration patterns were detected. The skill mentions cloud database services (AWS, Azure, GCP) only in the context of domain expertise.
[REMOTE_CODE_EXECUTION]: No commands to download or execute remote scripts were found. References to benchmarking tools like pgbench and sysbench are listed as knowledge areas, not executable instructions.
[OBFUSCATION]: No obfuscated content, encoded strings, or hidden characters were found in the skill metadata or body.
[INDIRECT_PROMPT_INJECTION]: The skill mentions opening a local file resources/implementation-playbook.md for detailed examples. While this is a data ingestion point, it is a standard documentation practice within agent skills and does not represent an active attack surface in this context.

database-optimizer