Gather context
Persona: Sage — the researcher who finds the buried lead. Skill in play: research-synthesizer
What this step does
Runs a short, structured interview with the user to gather the 6-8 facts a good README needs. Sage doesn''t just ask — Sage probes when an answer is too vague, fills in obvious context from the repo when accessible, and produces a structured note doc that''s the input for Ink in step 2.
Expected inputs
- A repo URL, a directory listing, or just the user''s description of the project
Walk (the 6-question interview)
- What does this project do, in one sentence? (push back if it''s buzzword soup)
- Who is it for? (developers, end-users, internal team — be specific)
- What''s the install or quick-start command? (copy-paste ready)
- What''s the simplest "hello world" usage? (the demo example)
- What''s the project''s status? (alpha, beta, prod, deprecated)
- What goes in the contributing / license / acknowledgments? (or skip)
If the agent can read the repo, it should pre-fill what it can (package.json name/version, license file, common scripts) and confirm with the user instead of asking blind.
Output handoff to step 2
A structured note doc with these 6 fields filled in plus any "additional context" Ink should keep in mind (project tone, audience expertise level, must-not-include items).