Status

blueprint status is the terminal view of Blueprint and Git state. It reads the project without changing either one. The same optional global command also provides the separate live dashboard.

Run the global command

If you installed the optional global command, use:

blueprint status

The global command does not install or update Blueprint files. Those operations remain under npx create-ai-blueprint@latest.

Run through npx

You can run the same status implementation without a global installation:

npx create-ai-blueprint@latest status

Project discovery

Status searches upward from the current directory until it finds a Blueprint project. You can run it from the project root or a nested directory.

Use --target to inspect a different directory:

npx create-ai-blueprint@latest status --target /path/to/project
blueprint status --target /path/to/project

Running status outside a Blueprint project without a valid target exits with a clear error.

Human output

The default output groups information into Project, Progress, Git, Attention, and Next action sections. Interactive terminals receive restrained color. Piped output remains plain.

Status reports:

• Project configuration state, effective review execution, regular and Continuous quality-gate modes, and the configured Git landing path
• Build-plan completion and the next unchecked item
• Active feature, fix, or rollback progress
• Open findings and completion blockers
• Independent-review state, freshness, requested reviewer, and actual reviewer model
• Git branch, working-tree state, upstream distance, and latest commit
• Overview drift and malformed or placeholder plan warnings
• One deterministic next action

If Git status inspection fails, including when its output exceeds the command buffer, Status keeps the rest of the report available and marks Git information as unavailable instead of exiting.

Valid activity that has stopped updating remains visible as interrupted, but it is informational rather than a project-health warning. A saved resume command does not override failed verification or malformed review/findings evidence.

For active work, recorded verification failure blocks completion and routes to /implement; incomplete or missing verification routes finished steps to /check. Malformed review/findings records or unsafe findings paths route to /doctor. All checked steps alone are not proof that work is ready. Status reports persisted evidence; it does not run the checks itself.

When no feature, fix, or rollback is active, completion reports idle rather than presenting the absence of work as a blocker. Overview freshness ignores build-plan checkbox completion markers, so /complete can update roadmap progress without creating false drift. Any other project-plan or build-plan change still makes the overview stale.

With the default git.landing: "local-merge", Status warns when the active feature is not the next unchecked build-plan item. It suppresses only that order warning for pull-request landing, where isolated parallel branches may handle different unchecked features. Missing plan items, completed targets, and branch prefix mismatches still produce warnings.

JSON output

Use --json for scripts and local integrations:

npx create-ai-blueprint@latest status --json
blueprint status --json

The command prints one versioned JSON object and does not include terminal color codes. Its configuration object includes the validated effective values, whether they came from the project file or built-in defaults, and any parsing warning. The review object exposes the receipt state, freshness, target, requested reviewer, model, and execution, plus the actual reviewer, model, execution, and reviewer context when present.

Agent workflow status is different

$status in Codex and /status in Claude Code are project-local agent skills. Google Antigravity also uses /status. In GitHub Copilot or OpenCode, ask the tool to run the status skill. These forms ask an AI agent to interpret the project and recommend the next workflow action. The terminal blueprint status command is a deterministic local CLI report that does not require an agent conversation.

See Workflow Status for the agent skill.