The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes explicit defensive instructions for the agent to ignore any directives (such as "ignore previous instructions") found within the data being analyzed (e.g., log files, email headers, or malware artifacts). This is a security best practice to mitigate indirect prompt injection attacks originating from adversarial data sources.
[DATA_EXFILTRATION]: While the skill assists in handling incidents involving sensitive data (PII/PHI), it explicitly instructs the agent not to exfiltrate or include full credentials/keys in its output, recommending generic placeholders instead. It lacks tools for network communication.
[COMMAND_EXECUTION]: The skill defines a purely analytical role, explicitly stating that it should not execute commands or interact with production systems. Its allowed-tools (Read, Grep, Glob) are restricted to file system observation and search.
[SAFE]: The skill provides high-quality guidance for incident response and incorporates security-by-design principles to protect the agent from untrusted input.

ir-playbook