See what your codebase actually does — and what it doesn't.
Point a coding agent at this repository and get a Service Blueprint or a User Journey Map as one self-contained HTML file. Every claim carries the address it came from. Every hole the code leaves is marked as a gap, because the holes are the reason you drew the thing.
Live file, not a screenshot. Scroll it sideways; click a gap in the summary to jump to it.
Read a repository and you see everything that exists. You do not see the confirmation email nobody sends, the failure screen nobody built, the queue whose errors nobody surfaces. Absence has no syntax highlighting.
Laying a flow out as a grid changes that. Phases run across, lanes run down, and an empty square is suddenly a question somebody has to answer. That is the whole trick, and it is why the interesting cells in these artifacts are the red ones.
So the rule the agent follows is blunt: a cell it cannot back with a real route or a
real file:line is a gap, not a confident guess. A sparse artifact whose
holes are true beats a full one that reads well.
In the repository you want to map, say:
create a service blueprint of the checkout flow using github.com/craice/journey
The agent fetches SKILL.md
and the template, reads your code, and writes the artifact to
docs/artifacts/. Ask it to update the file later and it diffs the artifact
against the code, leaving anything you corrected by hand alone.
Download template.html, replace the JSON block inside it per
SCHEMA.md, and open
the file. No build, no install, no network — it renders offline from a double click.
curl -fsSL https://raw.githubusercontent.com/craice/journey/main/template.html -o blueprint.html