-
Notifications
You must be signed in to change notification settings - Fork 0
Home
b2o2i edited this page May 10, 2026
·
9 revisions
AI checking AI.
Cross generates research reports using up to 5 AI providers simultaneously (or more via model agents), then cross-checks every report against all others — a 5×5 fact-check matrix by default. Reports can be published to Discourse.
New user? Start here: Onboarding — set up your API keys and run your first report in minutes.
| Command | Description |
|---|---|
| st | Interactive menu — launch any command from a numbered list |
| st-admin | Settings manager: DEFAULT_AGENT, model overrides, TTS voice, editor |
| st-analyze | AI-powered analysis of cross-product fact-check data |
| st-bang | Parallel report generation — run all AIs at once |
| st-cat | Print story fields to stdout (pipe-friendly) |
| st-cross | Cross-product: generate N stories × fact-check with N AIs |
| st-domain | Interactive wizard: create a new Cross-Stones benchmark domain |
| st-edit | Open a story container in your editor |
| st-fact | Fact-check one story with one AI |
| st-fetch | Import external content (URL, tweet, clipboard, file) |
| st-find | Search across story containers and prompt files |
| st-fix | Improve a story using fact-check feedback |
| st-gen | Generate a single AI report |
| st-heatmap | Cross-product fact-check score heatmap |
| st-ls | List contents of a story container |
| st-man | Manual pages for Cross commands (this system) |
| st-merge | Synthesize multiple AI stories into one |
| st-new | Create a new prompt from a template |
| st-plot | Cross-product fact-check score plots |
| st-post Post a story to Discourse | |
| discourse-workflows Discourse posting workflow guide | |
| st-prep | Prepare a raw AI response for posting |
| st-read | Readability metrics for stories |
| st-rm | Remove items from a story container |
| st-speak | Render a story as spoken audio (MP3) |
| st-speed | AI performance and speed analysis |
| st-stones | Cross-Stones benchmark leaderboard |
| st-verdict | Verdict category bar chart |
| st-voice | Browse, download, and test Piper TTS voices |
- Onboarding — first-time setup, API keys, running your first report
- Three Stages — GATHER · VERIFY · INTERPRET — how every Cross tool fits together
- Showcase Workflows — three killer workflows: "Is this fake news?", "What's missing?", "What can I trust?"
-
Container Format — anatomy of
subject.json:data[],story[],fact[], timing - AI Providers — all 5 providers, models, free vs paid tiers
-
Multi-Model — run more than one model per provider via agents (
anthropic-opus+anthropic-sonnetside-by-side) -
Agents — what an agent is, naming rules, resolution order, and the
--agentflag (cross-st 0.10.0+) - Cross-Stones Benchmark — benchmark suite: scoring, domains, leaderboard
- FAQ — common questions and troubleshooting
pipx install cross-st # no TTS
pipx install "cross-st[tts]" # with text-to-speech
st-admin --setup # configure API keys → ~/.crossenvSource: github.com/b202i/cross-st