The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill describes functionality for just-in-time provisioning of components from remote sources including Git, OCI, S3, and HTTP.
Evidence: source: "github.com/cloudposse-terraform-components/aws-vpc.git?ref=1.450.0" in SKILL.md. Note: The resource originates from the vendor's official GitHub repository.
[EXTERNAL_DOWNLOADS]: The dependencies.tools feature allows Atmos to automatically install and inject runtime CLI tools at various configuration scopes.
Evidence: Description of dependencies.tools in SKILL.md states "Atmos auto-installs and injects them, so do not add a separate install step."
[COMMAND_EXECUTION]: Components can override the default executable command, which allows for the execution of arbitrary binaries in place of standard infrastructure tools.
Evidence: The command field in the component configuration table in SKILL.md allows users to "Override the executable (e.g., tofu instead of terraform)."
[INDIRECT_PROMPT_INJECTION]: The skill describes an architecture where the agent processes potentially untrusted data from stack manifests and remote state outputs, which are then interpolated into execution contexts.
Ingestion points: SKILL.md, references/examples.md (Stack manifests, vars, env, and !terraform.state functions).
Boundary markers: None mentioned; the documentation does not describe explicit delimiters for untrusted variable values.
Capability inventory: SKILL.md (Subprocess execution of terraform, helmfile, and packer via atmos commands).
Sanitization: Not specified; the documentation does not mention validation or escaping of stack variables before execution.
[DYNAMIC_EXECUTION]: The skill outlines several mechanisms for dynamic code and configuration generation.
Evidence: references/component-types.md describes Atmos generating backend.tf.json and providers_override.tf.json files at runtime.
Evidence: references/examples.md demonstrates the use of Go Templates (eks-{{ .flavor }}/cluster) to dynamically generate component configurations based on context.

atmos-components