The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill documents the use of Python expressions for custom property mappings in SAML and OIDC providers (found in references/saml.md and references/integrations.md). These expressions allow for dynamic code execution within the Authentik environment's sandbox for attribute manipulation.
[DATA_EXFILTRATION]: The documentation in references/blueprints.md describes the use of specialized YAML tags such as !File and !Env. These tags provide the capability to access and read local file contents and environment variables during the configuration and deployment of Authentik blueprints.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted external identity data from LDAP directories and SAML Identity Providers (documented in references/blueprints/ldap_sources.md and references/google-source.md). This establishes a surface for indirect prompt injection where malicious attributes provided by external sources could influence authentication and authorization decisions.
Ingestion points: LDAP synchronization tasks and SAML assertion processing.
Boundary markers: The documentation recommends specific user_matching_mode settings and structured identifiers to manage data ingestion.
Capability inventory: Authentik can execute Python property mappings and modify its internal user database based on the ingested content.
Sanitization: Input sanitization is handled by the core Authentik platform; the skill provides the configuration templates for these data flows.
[COMMAND_EXECUTION]: The skill provides numerous shell commands for managing Kubernetes environments, including helm upgrade and kubectl exec (found in references/deployment.md and references/blueprints/ldap_sources.md). It also includes instructions for enabling the web terminal feature in ArgoCD.

authentik