Make starting a commons lead from existing material to a useful first decision

Nobody has claimed this yet.

A new commons currently gives too little guidance about what to bring and what useful result to expect. The first dogfood case is establishing the Reason Commons project's own goal and throughput; the Research Group provides a second case involving existing reasoning.

## Work

- Add a recoverable empty-commons starting experience with plain-language prompts for a goal, problem, conflict, possible solution or intended work, plus an unsure/exploratory route.
- Support describing a situation and bringing source material through the existing import path. Make the selected prompt guidance rather than a restriction to one tree.
- Ask what the group wants to work out. When relevant, guide agreement on success and throughput without requiring every new participant to settle the whole group's measure.
- Show a small source-backed interpretation, missing information and one useful next question; preserve the contributor's correction and human approval boundaries.
- Finish the missing import identity-conflict choices and browser upload coverage before relying on repeated import. Include pasted text/Markdown for the repo-based dogfood material.

## Acceptance criteria

A participant can start from real material, correct a misinterpretation, review a bounded proposal and return to the resulting accepted state. The experience is reopenable and needs no LTP vocabulary. A repeated identical import adds no duplicate work; conflicting wording has an explicit human resolution. Missing throughput remains visibly unresolved rather than fabricated.

## Expected benefit and check

A participant reaches a useful first judgment with less facilitator explanation and correction work. Observe that journey and record time, assistance and interpretation errors; agree on the target before the session.

Use the app's existing import and decision machinery. A fuller generated tree is not itself the success criterion.

Start with the app's existing import path and decision machinery, then inspect browser upload coverage and the repo-based dogfood material, including pasted text/Markdown. Done means a participant can import real material, correct an interpretation, review and accept a bounded proposal, reopen it, and resolve duplicate or conflicting imports without inventing missing throughput.

Written by the indexing model from the issue text.