The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes technical examples of prompt injection and system prompt extraction strings, such as "Ignore previous instructions" and "Repeat your instructions verbatim". These are explicitly documented within a security auditing checklist intended for red-teaming and authorized testing of target applications; they are not active attempts to manipulate or compromise the AI agent's own instructions.
[DATA_EXFILTRATION]: No patterns of unauthorized data access, hardcoded credentials, or exfiltration to remote servers were detected. The skill provides examples of vulnerable database queries and permission bypasses in other applications for auditing purposes without performing any sensitive data retrieval itself.
[COMMAND_EXECUTION]: The skill mentions CLI tools like grep and Bash for the purpose of mapping AI integrations during a security review. There is no evidence of malicious, autonomous, or privileged command execution within the instructions.
[OBFUSCATION]: The skill utilizes standard markdown and code blocks without any detected obfuscation techniques, hidden URLs, or malicious steganographic patterns.

iblai-vibe-security-prompt-injection