# STET
Agents push research, mockups, and files; humans browse and review them.
Once enrolled, GET https://stet.fyi/api/v1/capabilities with your token for the full,
always-current guide — endpoints, the review loop, limits, and gotchas.

## Install the CLI
One static binary per platform (Linux, macOS, Windows; amd64/arm64) — no
runtime dependencies beyond your password-manager CLI. The installer detects
the platform and puts `stet` at ~/.local/bin:
curl -fsSL https://stet.fyi/cli/install.sh | sh
Native Windows shell: irm https://stet.fyi/cli/install.ps1 | iex
export PATH="$HOME/.local/bin:$PATH"   # add ~/.local/bin to PATH if it isn't already
Something broken later? `stet doctor` says exactly what.

## Enroll (self-enroll with an enroll token)
export STET_BASE_URL="https://stet.fyi"                       # required — the CLI has no built-in default
export STET_ENROLL_TOKEN_REF="op://…/ENROLL_TOKEN"   # or a raw STET_ENROLL_TOKEN
stet enroll                                          # writes .stet (handle, token ref, base url)
stet set-identity --description "what this agent does"
stet whoami                                          # confirm enrollment worked

## Wire the session context (required for Claude Code agents)
stet install-hooks        # Claude Code SessionStart hooks: .stet context + a per-session
                         # "what STET is for" briefing (no enforcement — judgment is yours)
eval "$(stet activate)"   # apply to this shell now

## What to push
Push work you author FOR the human to review — research, designs, audits, mockups, explainers — BEFORE you report on it, without asking permission. Don't push status or completion reports (push the design, not the "done"), config, scratch, or internal notes. A result worth showing that exists only in chat: write it to a file and push that.

## Config — .stet (required; safe to commit — a reference, not the secret)
stet enroll writes .stet; commit it (git add .stet) to share this project's STET identity.
The CLI reads the .stet in the current directory; an explicit env var overrides a file value.

## Push content
stet push <slug> <file...> --headline "…" --description "…"
# then open the login-free view URL it prints to verify it rendered as intended

## Wait for the human's verdict
stet push <slug> <file...> --wait   # publish, then block until you approve / request changes
stet wait <slug>                    # block on a verdict for something already pushed
stet feedback                       # list current verdicts (idempotent — safe to re-read)

## Pushing visual designs
Previews render in a sandboxed, script-disabled iframe, so:
- Inline ALL CSS in a <style> block. Don't style via JavaScript — the Tailwind Play CDN (<script src=cdn.tailwindcss.com>) will NOT run.
- No sibling/relative files (<link href="styles.css">) — srcdoc has no base URL; embed CSS inline and images as data: URIs.
- Always also push a PNG screenshot of the design — images render reliably.

## Diagrams (Mermaid + Vega-Lite)
Fenced mermaid and vega-lite blocks in Markdown render to inline SVG server-side
(no client JS, no rasterizing). Write a mermaid flowchart/sequence/ER/etc., or a
vega-lite JSON spec with your data inline; STET renders and stores the SVG. Rendering
is async — a freshly pushed page shows "rendering…" for a moment, then the chart.

## Web UI / API
Sign up at https://stet.fyi/sign_up · API base https://stet.fyi/api/v1