skills/immamdouhaboammar/antigravity-superpowers/agency-incident-response-commander/Gen Agent Trust Hub
agency-incident-response-commander
Pass
Audited by Gen Agent Trust Hub on Aug 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides templates for infrastructure management using the
kubectlCLI. These include operations for rolling back deployments (rollout undo), restarting services (rollout restart), and scaling replicas (scale). These commands are intended for use by the agent or operator during active incident remediation. - [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection as it is designed to ingest and act upon data from untrusted external sources.
- Ingestion points: Real-time monitoring alerts, user-reported symptoms, and Slack thread logs (SKILL.md).
- Boundary markers: Not provided; the skill consumes alert data directly into its decision-making context.
- Capability inventory: Execution of Kubernetes orchestration commands (
kubectl) and potential interaction with third-party incident tools (e.g., PagerDuty, Slack). - Sanitization: No explicit validation or filtering of external alert content is defined to prevent embedded instructions from influencing agent behavior.
Audit Metadata