Capture Project Vision

Adapted from davidondrej/skills brain-to-docs at commit 11dee2ebc2d045806b686ba0b57746f1e3d7e331 (MIT). Its fixed documentation layout and write-after-every-answer rule were replaced with repository discovery and explicit approval.

Extract the user's purpose, taste, constraints, principles, and durable decisions into concise project documentation. This is an explicit interview, not an automatic side effect of ordinary development. Do not shape or implement a specific feature here; hand concrete build questions to shape-design, which owns feature-level design.

Discover conventions

Read the current README.md, project instructions, documentation index, CONTEXT.md or CONTEXT-MAP.md, and existing decision records. Infer where this repository keeps vision and decisions. If no convention exists, propose locations before writing; do not silently create a fixed structure.

Interview loop

1. Ask a small round of independent, high-variety questions about audience, problem, experience, principles, constraints, success, and non-goals.
2. Recommend answers only when repository evidence supports them.
3. Let the user answer whichever questions are useful.
4. Summarize the new facts and propose exact edits.
5. Write only after approval.
6. Re-read the changed docs before the next round so concurrent edits and settled answers become the new baseline.
7. Continue until the user says the vision is sufficiently captured.