CLI reference

Run synthworld --help and the relevant subcommand --help for the installed version. Getting Started and Evaluating a system contain runnable examples; the root user guide is only a historical compatibility index.

The unified synthworld evaluate <task> command accepts agentic, broker, continuous-assurance, contextual-access, enterprise-agentic, generated-enterprise-agentic, extraction, entity-resolution, relationship, and risk. Some additional package surfaces remain Python-only; use the relevant contract rather than inferring a CLI.

Enterprise authorization currently has CLI commands only for scaffolding, validating, and compiling the identity/access import. Corpus construction, directory/RBAC, ABAC, ReBAC, composition, prediction, and scoring use the documented synthworld.enterprise.consumer Python API. See Build and score an enterprise authorization experiment.

Render the published Asteria authority world

Create a deterministic, self-contained public HTML view from a generated or reproduced Asteria Agentic v1 public package:

synthworld generate-agentic --output asteria-agentic-v1
synthworld visualize \
  --public-package asteria-agentic-v1/public \
  --view agent-authority \
  --output asteria-public.html

Reference truth is opt-in and must come from the separately verified evaluator tree:

synthworld visualize \
  --public-package asteria-agentic-v1/public \
  --evaluator-package asteria-agentic-v1/evaluator \
  --view agent-authority \
  --output asteria-evaluator.html

Evaluator output is visibly watermarked. The default --package-profile asteria accepts only the published Asteria v1 artifact-set digest and refuses to overwrite an existing file. It does not render enterprise authorization packages.

generate-enterprise-agentic preserves its fixed-reference default. Select a generated tier explicitly:

synthworld generate-enterprise-agentic \
  --profile generated \
  --tier smoke \
  --seed 20260814 \
  --output generated-enterprise-agentic

The generated choices are smoke, standard, and longitudinal. Smoke retains the released EnterpriseAgenticGenerationConfigV1 contract. Standard and longitudinal use the independently versioned EnterpriseAgenticGenerationConfigV2 family. The command writes separate public/ and evaluator/ trees and refuses to replace an existing output root. Pass a complete JSON configuration with --config; explicit --tier and --seed values override those two resolved fields. --public-only writes no evaluator directory. See Generate enterprise-agentic scale tiers.

Render a generated enterprise-agentic world

Select the generated package contract explicitly to render a verified generated public tree with the same shared Explorer renderer:

synthworld visualize \
  --public-package generated-enterprise-agentic/public \
  --view agent-authority \
  --package-profile generated-enterprise-agentic \
  --output generated-public.html

The public command reads only the public tree and never opens evaluator truth. Passing --evaluator-package generated-enterprise-agentic/evaluator as well verifies both trees, their digest cross-binding, and declared generator conformance before producing watermarked evaluator output. Unsupported generated tiers or package versions fail explicitly rather than rendering a partial view. Layout coordinates are computed deterministically from the projection alone, so the same verified package always produces the same HTML bytes.

The fixed and generated enterprise-agentic submissions are deliberately separate CLI tasks because they use different trace contracts. Reload and check a generated observed-action trace using only its public tree:

synthworld validate generated-enterprise-agentic-trace \
  --benchmark-root generated-enterprise-agentic \
  --predictions observed-actions.jsonl

Score it only after the complete root is available to the evaluator process:

synthworld evaluate generated-enterprise-agentic \
  --benchmark-root generated-enterprise-agentic \
  --predictions observed-actions.jsonl \
  --summary

Validation checks structure and action-event coverage; it does not claim the reported observations are correct. The generated benchmark guide defines the required event replay and public/evaluator separation.

See Run an enterprise agentic identity experiment for an end-to-end RBAC/ABAC/ReBAC pilot and explicit public projection, layout, evaluator-overlay, and HTML generation steps.