Rust CLI
Automate the complete pipeline.
Score, rewrite, train a strategy policy, benchmark quality, verify corpora, or commission a human rewrite. JSON and report output preserve the evidence.
Read the CLI reference →Detector-directed rewriting
Skryba generates alternatives, measures every one against the detectors you configure, and rejects anything that fails its quality judge.
$ skryba rewrite draft.txt --report report.json
chunk 1/3
generated 8 candidates
quality 6 passed
selected plainspoken
ai probability 0.14
chunk 2/3
source kept no candidate passed qualityOne pipeline, visible evidence
Skryba does not equate a lower detector score with a better sentence. Detectability and quality stay separate until selection.
Long input becomes bounded chunks, keeping paragraph boundaries where possible.
A learned policy samples editing strategies and Brama generates multiple candidates.
Commercial detectors estimate AI probability while an LLM judge checks meaning, facts, grammar, coherence, naturalness, and readability.
Candidates with critical errors or low quality are rejected. The least detectable survivor wins; otherwise Skryba keeps the source.
Work where you work
Rust CLI
Score, rewrite, train a strategy policy, benchmark quality, verify corpora, or commission a human rewrite. JSON and report output preserve the evidence.
Read the CLI reference →macOS desktop
Score, Rewrite, and RentAHuman Status each show the exact command, carry stdout and failures across unchanged, and keep credentials below the view.
Explore Skryba Desktop →Guardrails, not guesses
If no generated candidate preserves meaning above the configured threshold, Skryba keeps the original chunk. It never needs a bad rewrite to finish a run.
Complete documentation
The website carries the full CLI and desktop corpus: concepts, references, walkthroughs, architecture, configuration, and runbooks.
Browse all documentation