/review
Inspect Wakeflow target review inputs, independently validate behavior, and record an explicit controller decision
$ npx -y skills add GxFn/Wakeflow --agent claude-codeHow it fires
How this command gets triggered: by you, by Claude, or both.
- Fires itselfClaude auto-loads it when your prompt matches the work.
- You can call itInvoke it directly when you want it.
- Slash command
/review
Context preview
What this command does when you run it.
Inspect Wakeflow target review inputs, independently validate behavior, and record an explicit controller decision
Command definition
review.mddescription: Inspect Wakeflow target review inputs, independently validate behavior, and record an explicit controller decision
argument-hint: [dispatch-group-or-task-id]
Run one controller review step. Use the wakeflow-controller skill acceptance rules; target backfill is input, never a conclusion.
1. Confirm this session is the controller window. Resolve the dispatch group or task id from `$ARGUMENTS`, or from the newest result-ready group in `wakeflow_status`. 2. Call `wakeflow_review_pack` for the group/task and inspect the listed target-authored inputs (files, commits, commands, logs, reports), then independently validate the relevant behavior before judging anything. Each entry also carries the intent triple — Design's `designIntent`, the authored dispatch `objective`, and the result — plus an `intentCheck` reminder; it is advisory context for YOUR drift judgment (no scores, no gates), and a real drift routes to redesign, not a point fix. 3. When multiple target results need consolidation, call `wakeflow_reduce_results`; treat the reduction as a review candidate only. 4. Decide explicitly and record it: `wakeflow_decide_review` with accept / rework / blocked / redesign, or `wakeflow_complete_demand` when the completion definition is met with evidence. Mainline redesign uses the stateless mainline Design handoff before a replacement task. A Pod must not fall back to mainline Design; the current implementation supports only one frozen Pod Design request/handoff generation (whose sole request may be `initial-design`, `supplement`, or `redesign`), so a different second request stays blocked as a capability gap. 5. After recording the decision, clear transient window glyphs with `wakeflow-claude-host window-status --reconcile --root <workspace>` (green + means result-ready; reconcile returns settled windows to plain tabs). 6. Report the decision, the evidence that supports it, and the next eligible step (next package, intake, archive, or stop for user judgment).
Never accept work merely because target material shows that a script ran. Independently inspect or reproduce the intended behavior. When review inputs are missing or ambiguous, choose rework or blocked and say why; when the work is valid but the outcome misses the requirement, choose redesign.
Read more
description: Inspect Wakeflow target review inputs, independently validate behavior, and record an explicit controller decision argument-hint: [dispatch-group-or-task-id]
Run one controller review step. Use the wakeflow-controller skill acceptance rules; target backfill is input, never a conclusion.
1. Confirm this session is the controller window. Resolve the dispatch group or task id from `$ARGUMENTS`, or from the newest result-ready group in `wakeflow_status`. 2. Call `wakeflow_review_pack` for the group/task and inspect the listed target-authored inputs (files, commits, commands, logs, reports), then independently validate the relevant behavior before judging anything. Each entry also carries the intent triple — Design's `designIntent`, the authored dispatch `objective`, and the result — plus an `intentCheck` reminder; it is advisory context for YOUR drift judgment (no scores, no gates), and a real drift routes to redesign, not a point fix. 3. When multiple target results need consolidation, call `wakeflow_reduce_results`; treat the reduction as a review candidate only. 4. Decide explicitly and record it: `wakeflow_decide_review` with accept / rework / blocked / redesign, or `wakeflow_complete_demand` when the completion definition is met with evidence. Mainline redesign uses the stateless mainline Design handoff before a replacement task. A Pod must not fall back to mainline Design; the current implementation supports only one frozen Pod Design request/handoff generation (whose sole request may be `initial-design`, `supplement`, or `redesign`), so a different second request stays blocked as a capability gap. 5. After recording the decision, clear transient window glyphs with `wakeflow-claude-host window-status --reconcile --root <workspace>` (green + means result-ready; reconcile returns settled windows to plain tabs). 6. Report the decision, the evidence that supports it, and the next eligible step (next package, intake, archive, or stop for user judgment).
Never accept work merely because target material shows that a script ran. Independently inspect or reproduce the intended behavior. When review inputs are missing or ambiguous, choose rework or blocked and say why; when the work is valid but the outcome misses the requirement, choose redesign.
A disciplined control loop for multi-window agent work — every step traced, every result reviewable.
Repo: GxFn/Wakeflow
Other commands on wakeflow.
- /check
Health-check an existing Wakeflow workspace and converge missing or outdated Claude Code surfaces
Open command - /dispatch
Prepare and send one Wakeflow delivery to a target window session (controller only)
Open command - /init
Initialize the current workspace as a Wakeflow control workspace (dry-run first, apply after confirmation)
Open command - /status
Show Wakeflow workspace status — demands, eligible work, deliveries, and window readiness
Open command - /unattended
Turn unattended (prompt-free) mode for the work windows on or off, then reopen them
Open command - /windows
List, launch, resume, or attach Wakeflow tmux windows, choosing the permission mode
Open command