The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to the way it processes external data.
Ingestion points: The skill frequently ingests untrusted data via the $ARGUMENTS placeholder in almost every phase (Detection, Triage, Investigation, etc.) within SKILL.md.
Boundary markers: There are no boundary markers or instructions provided to the sub-agents to distinguish between the workflow instructions and potentially malicious content within the alert/incident data.
Capability inventory: The skill orchestrates agents with high-impact capabilities including rollback assessment, traffic throttling/rerouting, emergency deployment, and production fix implementation (Steps 3, 7, and 8 in SKILL.md).
Sanitization: No sanitization, validation, or escaping of the $ARGUMENTS content is defined before it is interpolated into agent prompts.
[COMMAND_EXECUTION]: The skill explicitly instructs agents to perform high-privilege operations.
Evidence: Steps 7 and 8 in SKILL.md involve 'Design and implement production fix' and 'Execute emergency deployment'. While these are the primary purpose of an incident response skill, they represent a high-risk capability if triggered by a malicious indirect injection.

incident-response-incident-response