The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill configures a reverse proxy (Traefik) that ingests untrusted data from the network (Host, Path, and Headers matchers) which could potentially influence agent behavior if processed without adequate sanitization.
Ingestion points: references/routing.md defines matchers for Host(), Path(), Header(), and Query() parameters.
Boundary markers: Not present in the provided YAML configuration examples.
Capability inventory: SKILL.md includes capabilities to execute helm install and kubectl apply commands.
Sanitization: No specific sanitization or filtering logic is provided for the interpolated request data in the examples.
[CREDENTIALS_UNSAFE]: references/kubernetes-crd.md contains a Kubernetes Secret example with a hardcoded password password123. While likely intended as a generic placeholder, it is a hardcoded secret in the instruction set.
[EXTERNAL_DOWNLOADS]: SKILL.md instructs the agent to download Helm charts from https://traefik.github.io/charts and apply CRD definitions from Traefik's official GitHub repository (raw.githubusercontent.com/traefik/traefik). These are neutral references to the official resources of a well-known service provider.

traefik