Summary

Adds a focused manuscript-certainty-tone-assistant/ slice for the AI-Assisted Research Tools MVP in #13.

The assistant calibrates manuscript claim wording against linked evidence strength. It flags overconfident causal or definitive language, universal scope without replication, broad population wording from small samples, unsupported significance language, missing uncertainty intervals, missing hedging, and missing limitation anchors.

Distinctness

This is intentionally scoped to certainty and tone calibration. It is not another broad research tools suite, evidence summarizer, citation metadata/context/diversity/style/retraction checker, manuscript similarity triage, ethics/data availability auditor, unit consistency checker, biomethods provenance assistant, image integrity check, or broad statistical consistency module.

Safety

• Synthetic fixtures only
• No external API calls
• No credentials, tokens, private manuscripts, live DOI records, PubMed/arXiv/Crossref lookups, publisher data, or payment data
• Deterministic reviewer packet digests

Validation

• npm run check
• npm test
• npm run demo
• git diff --check
• git diff --cached --check
• ffprobe -v error -select_streams v:0 -show_entries stream=codec_name,width,height,duration,avg_frame_rate -show_entries format=size,duration -of default=noprint_wrappers=1 manuscript-certainty-tone-assistant/reports/demo.mp4 -> H.264, 960x540, 5s, 15 fps

/claim #13