The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to review user-provided Kubernetes manifests, which are untrusted external inputs. These manifests could contain malicious instructions embedded in comments or metadata designed to influence the agent.\n
Ingestion points: The workflow step 'reviewing existing manifests' in SKILL.md and the pre-deployment checklist involve reading external YAML content.\n
Boundary markers: The skill does not provide specific instructions to the agent on how to isolate data from instructions when parsing manifests.\n
Capability inventory: The skill mentions tool usage such as kubectl and kubeconform.\n
Sanitization: There is no explicit instruction to sanitize manifest content before the agent evaluates it.\n- [CREDENTIALS_UNSAFE]: The Kustomize deep dive in references/kustomize.md contains demonstration credentials (username=admin, password=secret123) within a secretGenerator literal block.\n- [EXTERNAL_DOWNLOADS]: The skill documentation includes examples of Kustomize remote bases that fetch configuration from GitHub repositories.\n- [COMMAND_EXECUTION]: The skill instructs the agent to use command-line tools like kubectl and kubeconform to apply and validate resources within the Kubernetes cluster.

kubernetes-manifests