Track progress

Three places tell you where a spec is: the rail, the sidebar and the footer. You don’t have to do anything to keep them current. They follow the files the run writes.

Place	What it shows
The rail	A check when a step’s document exists, a spinner and a timer while it runs, and a lock on documents a running step has not reached
The Tasks tab	During implement, a live percentage from the checked boxes in tasks.md
The sidebar	For a spec in flight, its last finished task and how long ago, like T004 · 2h ago
The footer	The next step by name, with Next: Plan above the button. No forward button while a step runs.
A notification	When a step or a phase of tasks completes, with Open spec. Turn it off with speckit.notifications.stepComplete.

A run in the viewer: the rail shows which documents exist and the footer names the next step.

While a step runs, its tab spins and keeps time.

Three commands for three questions

All three need the Companion Spec Kit extension and work with any assistant.

Question	Command	What it does
Where does this spec stand?	/speckit.companion.status	Reports the current step, its status, the decisions recorded so far and the next action. Changes nothing.
I stopped halfway. How do I carry on?	/speckit.companion.resume	Finds the next step, carries the recorded decisions into scope and sends that step’s command. In implement it continues at the next unchecked task.
Did the run do what it says it did?	/speckit.companion.doctor	Recomputes what happened and reports problems. Read-only, and it never stops your work.

Resume from the sidebar

Hover a spec that is still in flight and click the play button. That is Resume. It shows only when the spec can move forward and the Companion Spec Kit extension is installed. On a spec that is already implemented, completed or archived, resume reports that the run is complete and sends nothing.

Hover a spec to get Resume and More Actions.

Run the doctor

/speckit.companion.doctorCopied

Add	To
Nothing	Check the active spec
--feature-dir specs/042-export-csv	Check one spec by name
--all	Check every spec
--json	Get a machine-readable report
--chat	Also read the assistant’s session transcript, and explain failed or abandoned work

It looks for the things a run can get wrong without anyone noticing.

It finds	For example
Unfinished work	A step that started and never finished, or a task ticked in tasks.md with nothing recorded for it
A step doing the next step’s work	Plan content in the spec, a task list in the plan, code written before implement
A step that closed having verified nothing	Implement finished but no test or check ever ran
A missing file	A step that closed without the document it promised
Why a spec did not complete	The write was refused, it never arrived, or it landed and the display disagrees

Every check says whether it ran, was skipped (with the reason) or does not apply, so a check that could not look never reads as clean. The doctor reports and stops. Fixing what it finds is your call.

Every capture call writes one line to specs/<NNN>/.trace.jsonl, which is what the doctor reads to explain a failed capture. The file is safe to delete.

Related

The Overview shows what a finished run recorded.