CLI + macOS desktop
Skryba documentation
Learn the rewrite pipeline from first score to selected text, then inspect every command, concept, workflow, trust boundary, and failure mode.
Start here
CLI reference
CLI referenceCommands, global options, input rules, and output contracts.HandToHuman and RentAHumanCommission and inspect a human rewrite from the command line.Train, benchmark, and corpusTrain strategy policy, benchmark quality, and verify curated samples.ConfigurationModels, detectors, credentials, cache, and endpoint overrides.ExamplesRunnable examples shipped with the CLI repository.CLI runbookSymptom-first diagnosis for model, detector, cache, and workflow failures.
Concepts
RewriteChunked candidate generation and selection.ScoreEnsemble AI probability and detector evidence.ChunkHow long inputs become bounded rewrite units.DetectorProvider adapters, score normalization, and ensemble behavior.Quality judgeMeaning preservation, critical errors, and thresholding.Detector cacheRevision-aware score reuse and billing protection.Strategy policySampling, GRPO training, and policy files.Credential boundaryWhich component receives each credential.Humanization orderThe durable boundary for commissioned human work.
Walkthroughs
Desktop
What is Skryba Desktop?The native macOS process client and its mental model.Desktop quick startInstall the CLI, configure Finder launches, and run the app.Desktop Score workflowInput, invocation, result, and failure behavior.Desktop Rewrite workflowControls and rendered candidate evidence.Desktop RentAHuman StatusFetch and display one owned humanization.Desktop spawn contractExecutable search, process I/O, and compatibility.Desktop configurationEnvironment merge rules and build-time variables.Desktop runbookSeparate form, spawn, CLI, and decoding failures.Desktop architectureSwiftUI components and the CLI trust boundary.