The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill is designed to capture and extract sensitive information, including Secrets, Tokens, and Flags, from Kubernetes Mutating Webhooks by forging AdmissionReview requests.
Evidence: Phase 3 in SKILL.md demonstrates extracting Base64 patches containing values like wiz_k8s_lan_party{...}.
[DATA_EXFILTRATION]: Forging AdmissionReview requests to trick internal cluster services into revealing sensitive configuration data intended for other resources.
Evidence: Usage of curl to POST fake requests to internal service endpoints such as https://kyverno-svc.kyverno.svc.cluster.local/mutate.
[COMMAND_EXECUTION]: Utilizes system commands to perform network operations, file creation, and data processing.
Evidence: Usage of curl, cat, base64, and jq to interact with Webhooks and parse stolen data.
[PRIVILEGE_ESCALATION]: The techniques facilitate bypassing admission control security mechanisms to obtain privileged credentials, enabling lateral movement within a Kubernetes environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes responses from external services (Webhooks) without sufficient sanitization, creating an attack surface for indirect injection if the service is compromised.
Ingestion points: Responses from curl commands in SKILL.md Phase 2 and Phase 4.
Boundary markers: Absent; response data is piped directly to processing tools.
Capability inventory: Network operations (curl), file writing (cat), and data parsing (jq, base64).
Sanitization: Absent; the skill directly decodes and parses responses from the target service.

k8s-webhook-abuse