The job
You have a PRD open in one Notion tab and interview notes open in another, and you need a direct map between proposed features and what users actually asked for. You want a clean list of matches, conflicts, and missing requirements before the next planning review, without rebuilding both docs by hand in a separate chat window.
Why this is hard without Sephir
Notion AI reads one page at a time, so cross-document analysis turns into manual reconstruction. You end up copy-pasting two long pages into another tool, stripping callouts and headings, then re-prompting to recover structure each time the answer drifts. After a few turns, the context is bloated with boilerplate and the important signal gets buried in repeated text.
How Sephir does it
- Pin the two Notion tabs you want to compare.
- Open Sephir in the sidepanel with
Cmd+Shift+S. - Ask for a feature-to-feedback gap map with conflicts and missing items.
- Watch Sephir run
extractPageText(tab 1)andextractPageText(tab 2)in the audit timeline. - Review the synthesized output grouped as aligned, contradictory, and absent requirements.
- Save or export the result as Markdown for your planning doc.
The skill behind it
This skill runs the same repeatable flow each time: read both tabs, compare them directly, and return a structured synthesis you can act on in product planning.
What it costs
Sephir runs this on your own provider setup. If you use ChatGPT Plus via Codex OAuth, this workflow adds zero extra token charges. With a direct API key, expect about ~4,000–7,000 input tokens and ~700–1,200 output tokens per run on Claude Opus 4, GPT-5.5, or Gemini 3 Pro. License details are on .