The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data, including Infrastructure-as-Code (IaC) definitions (Terraform, CloudFormation, Pulumi), resource utilization metrics, and business data from Pulse. This creates an attack surface for indirect prompt injection where instructions embedded in metadata or resource names could potentially influence the agent's logic.
Ingestion points: IaC source code, plan outputs, cloud utilization metrics (CPU, memory, GPU occupancy), and business KPIs.
Boundary markers: The instructions lack explicit delimiters or specific 'ignore-instructions' directives to isolate ingested data from the agent's core reasoning instructions.
Capability inventory: The skill has the capability to execute CLI tools (infracost, jq, bc) via subprocess calls, write reports to the file system, and perform network requests for pricing data.
Sanitization: The skill relies on external CLI tools for data parsing and does not detail internal sanitization of the values extracted from the data before they are processed by the LLM.
[EXTERNAL_DOWNLOADS]: The skill utilizes Infracost, a well-known third-party CLI tool for cloud cost estimation, and references various reputable external documentation and research sources for AI cost optimization (e.g., projectdiscovery.io, martinfowler.com). While these are standard in the engineering domain, they represent external dependencies.
[COMMAND_EXECUTION]: The skill provides recipes and instructions for executing shell commands and CLI tools (infracost, jq, bc) as part of CI/CD gates and cost analysis workflows. These operations are conducted on technical files (IaC plans) and are central to the skill's utility, requiring the agent to handle external file paths and configurations.

ledger