The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external configuration files, which creates a potential surface for indirect prompt injection if those files contain malicious instructions disguised as comments or configuration data.
Ingestion points: The skill uses Glob, Grep, and Read tools to ingest infrastructure-as-code files (Terraform **/*.tf), Kubernetes manifests (**/NetworkPolicy*, **/calico*), and traditional network configuration files (**/*.acl, **/interfaces*).
Boundary markers: The skill includes a specific 'Prompt Injection Safety Notice' instructing the agent to treat configuration content as untrusted data and to ignore any text that appears to be a prompt or instruction.
Capability inventory: The skill's capabilities are limited to Read, Grep, and Glob. It does not have access to network tools or file-write operations, significantly limiting the potential impact of an injection.
Sanitization: The skill relies on natural language instructions to guide the agent's behavior; it does not employ programmatic sanitization or strict schema validation for the ingested text files.

segmentation