@vitra/cli — vitra

CLI for the Vitra Image Creator. Same backend, same tool surface, and same credential as the MCP server (@vitra/image-creator-mcp) and the SKILL.md drop-in. Use the CLI for headless scripting, CI, or just-the-terminal.

Spec: docs/superpowers/specs/2026-05-12-animoca-skill-pack-v4-canonical.md.

Install

npm i -g @vitra/cli
vitra login                       # paste your vitra_sk_live_* secret
vitra doctor                      # verify credential, channel, manifest

Commands

vitra login | logout
vitra config get | set <key> [value] [--push|--pull]
vitra doctor [--managed-equivalent]
vitra brand-kit create --url URL
vitra brand-kit list
vitra brand-kit get ID
vitra brand-kit update ID [--name "..."] [--voice "..."] [...]
vitra brand-kit delete ID
vitra brand-kit versions ID
vitra image generate [--brief "..."] [--headline "..."] [--cta "..."] [--channel ID] [--enrich | --no-enrich]
vitra image edit --asset-id ID --instruction "..."
vitra image translate --asset-id ID --locale BCP47
vitra image resize --asset-id ID [--channel ID | --width PX --height PX | --aspect-ratio W:H]
vitra webhooks register URL --events LIST
vitra webhooks list | delete ID

audit and dev (sandbox) are on the v1 plan.

Brief enrichment — --enrich

vitra image generate --enrich runs an optional pre-generate step that expands a sparse brief into a structured, provider-tuned prompt — without deviating from intent. Numbers, percentages, prices, dates, and quoted / --headline / --cta text are reproduced verbatim; enrichment only adds composition, lighting, camera, and style. It costs one extra cheap text-model call (~$0.0004, +1–3s). After the render the CLI prints a plain-text "why this prompt" receipt — locked items vs. what was added; on failure it falls through to the raw brief (generation never blocks).

Requires the tenant's brief_enrichment feature (enabled in the admin panel). --no-enrich forces it off for one call; omit both flags to use the account default.

Credential

~/.vitra/credentials.json (mode 0600). Env override: VITRA_HMAC_SECRET.

vitra config syncs ~/.vitra/config.json with the server-side config store at GET /api/skill/config — --push writes local → server, --pull writes server → local. This is how a tenant admin can set auto_approve_when policies once and have every developer's CLI/MCP pick them up.

--json flag

Every command supports --json for scripting:

vitra brand-kit list --json | jq '.data[] | select(.name == "Animoca Core")'

When --json is set, the CLI does NOT prompt interactively — it returns the needs_clarification envelope as JSON and exits. Useful for shell pipelines and CI.

Interactive prompts

Without --json, the CLI handles needs_clarification envelopes via the prompts library: numbered select, multi-checkbox, text input with validation, etc. Render contracts per each of the 10 clarification kinds land in MR-2 and MR-4.

License

(TBD — set during MR-12 GA cutover.)