The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill documents methods to extract highly sensitive cryptographic material, including TLS private keys (tls.key, key.pem) and CA certificates from Kubernetes secrets and internal Pod directories (/var/run/secrets/istio/). This allows for service identity spoofing and unauthorized decryption of mesh traffic.
[DATA_EXFILTRATION]: Instructions are provided to redirect or mirror all service traffic to an external host (attacker-service) using VirtualService configurations. Additionally, the skill guides the user to scrape service logs for sensitive credentials such as passwords, tokens, and session cookies.
[COMMAND_EXECUTION]: Extensive use of shell-based orchestration (kubectl, curl) is employed to interact with the Kubernetes API and local Envoy Admin interfaces (localhost:15000) to enumerate mesh topology and manipulate the operational state of the cluster.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) as it ingests and processes untrusted data from external sources like Kubernetes logs and Envoy configuration dumps without implementing boundary markers or sanitization, while maintaining high-impact cluster-write capabilities.

servicemesh-tactics