The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documents the atmos auth exec and atmos auth shell commands, which allow for the execution of arbitrary shell commands and interactive sessions with injected cloud credentials.
[INDIRECT_PROMPT_INJECTION]: The skill defines a surface for processing external authentication material (OIDC tokens, SAML responses, and SSO device codes) that dictates the security context for subsequent agent actions.
Ingestion points: OIDC tokens from GitHub Actions, GCP Workload Identity Federation tokens, and SAML assertions from identity providers like Okta or Google Apps.
Boundary markers: No specific delimiters for prompt-level token handling are defined; validation and exchange are handled by the underlying atmos CLI utility.
Capability inventory: Authentication network requests, writing configuration files (kubeconfig, Docker config) to standard XDG paths, and arbitrary command execution via atmos auth exec.
Sanitization: The skill relies on standard industry-standard authentication protocols (OpenID Connect, SAML 2.0, AWS SSO) to validate and exchange tokens.
[SAFE]: The skill demonstrates safe secret management by instructing users to use the !env YAML tag for sensitive credentials and recommending OS-native secure storage (keyrings) for cached session data.

atmos-auth