CLI reference

Generated from the argument parser; python cli.py <command> --help prints the same text.

usage: dokimi [-h] [--version]
              {run,evaluate,compare,baseline,report,personas,questions,answers,eval,library,product,benchmark,dashboard,arms,subjects,tasks,catalog,explore,dataset,ingest,novelty,repo,tools,context} ...

positional arguments:
  {run,evaluate,compare,baseline,report,personas,questions,answers,eval,library,product,benchmark,dashboard,arms,subjects,tasks,catalog,explore,dataset,ingest,novelty,repo,tools,context}
    evaluate            Run full evaluation pipeline (one command)
    baseline            Save and compare eval baselines
    report              Generate analysis reports
    personas            Persona discovery and management
    questions           Question generation
    answers             Answer generation
    eval                Evaluate answers
    library (product)   Product registry management
    benchmark           Run context benchmark for one or more registered targets
    dashboard           Generate dashboard from evaluation results
    arms                Compare context-augmentation treatments (docs, MCP, skills, agent
                        profiles)
    subjects            Run evaluation subjects (artifact + suite) into scorecards
    tasks               Run executable coding tasks through Codex, Claude Code, or terminal-bench
                        harnesses
    catalog             Scan an external skill/task catalog (e.g. a local uxlfoundation/skills or
                        intel/skills checkout)
    explore             Run exploration-only localization tasks (ExploreBench)
    dataset             Export artifacts to shareable datasets
    ingest              Materialize Markdown from other formats
    novelty             Measure what an artifact's content adds to a model
    repo                Repositories as evaluation subjects: validate repo-eval.yaml, L0 scan,
                        fetch sources, reach probe, task self-check, agent runs and their report
    tools               How a model chooses among the tools it is offered
    context             What an agent's context window held during a run, read from its trace. No
                        LLM call.

options:
  -h, --help            show this help message and exit
  --version             show program's version number and exit