nextaltair/video-library-pipeline-inventory-review

SourceRoot V2 Run Start

Rule

• Use only the V2 public tools.
• Start with video_pipeline_start; do not call hidden legacy tools.
• Do not infer latest files from windowsOpsRoot/llm or windowsOpsRoot/move.
• Stop for human review whenever the returned nextActions or followUpToolCalls require it.

Tool Sequence

1. Call:

   video_pipeline_start {
     "flow": "source_root"
   }
   

   Optional parameters:
   • runId when the operator needs a stable explicit ID
   • maxFilesPerRun
   • allowNeedsReview
   • driveRoutesPath
2. Read WorkflowResult:
   • runId
   • phase
   • outcome
   • artifacts
   • gates
   • nextActions
   • followUpToolCalls
   • diagnostics
3. If phase == "review_required", hand off to skills/extract-review/SKILL.md with the full result.
4. If phase == "plan_ready", hand off to skills/move-review/SKILL.md with the returned plan action.
5. If phase == "blocked" or phase == "failed", report diagnostics and stop.

Human Review Checklist

• Confirm the runId is included in every follow-up.
• Confirm all referenced artifacts belong to the same run.
• If a ReviewGate is open, inspect its artifact IDs before resuming.
• Do not continue from free-text nextStep; use structured nextActions or followUpToolCalls.

Handoff

• To metadata review: pass runId, open gate details, and artifact IDs from gates.
• To move review: pass the followUpToolCalls entry whose reason/action is plan review.