The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides instructions for analyzing infrastructure command outputs (e.g., from Terraform or Atmos stacks) using AI agents via the --ai flag or atmos ai ask. This creates an attack surface where untrusted data within the stack output could attempt to inject malicious instructions into the agent's context.
Ingestion points: Output from commands like atmos terraform plan, atmos describe, and user-provided queries in atmos ai chat.
Boundary markers: Not specified in the documentation.
Capability inventory: The skill allows for extensive CLI command execution and tool orchestration across the infrastructure lifecycle.
Sanitization: Not explicitly mentioned in the skill instructions.
[EXTERNAL_DOWNLOADS]: The documentation describes the atmos ai skill install command, which allows users to download and install logic from arbitrary external GitHub repositories. While a core feature of the Atmos ecosystem, users must verify the provenance of any third-party skills.
[COMMAND_EXECUTION]: The skill defines and uses a variety of CLI commands to manage cloud infrastructure, handle authentication via atmos auth, and start MCP servers for AI integration. This includes a 'yolo' mode which, according to the documentation, bypasses safety checks for tool approval.

atmos-ai