The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains explicit safety instructions in Section 8 (Prompt Injection Safety Notice) directing the agent to ignore any 'ignore previous instructions' or similar directives found within analyzed incident artifacts. These are defensive controls designed to protect the agent's integrity when processing untrusted data, rather than malicious injection attempts.- [DATA_EXFILTRATION]: While the skill processes potentially sensitive data (logs and evidence), it instructs the agent to minimize data exposure by referencing sensitive values generically rather than exfiltrating raw secrets or credentials.- [EXTERNAL_DOWNLOADS]: The skill references several external resources for incident response frameworks. All referenced domains (nist.gov, sans.org, cisa.gov, sec.gov, mitre.org, etc.) belong to trusted government organizations or well-known cybersecurity entities.

ir-playbook