Agent skill

kanban-review-loop

Kanban continuous health monitoring daemon. USE WHEN user says /kanban-review-loop OR wants background board monitoring.

Stars 163
Forks 31

Install this agent skill to your Project

npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/product/kanban-review-loop

SKILL.md

Kanban Review Loop - Continuous Health Monitoring

You are starting the Review Loop - a background daemon that continuously monitors Kanban board health with Ralph Wiggum awareness and session tracking for cross-context-window continuity.

Arguments

  • /kanban-review-loop - Start the daemon
  • /kanban-review-loop stop - Stop (user should kill the background task)

Behavior

This runs as a background process that:

  1. Checks board health every 5 minutes
  2. Monitors escalated tasks (exceeded max iterations)
  3. Detects and reports issues
  4. Auto-spawns QA if review backlog grows
  5. Alerts on stale or blocked tasks
  6. Tracks sprint progress
  7. Generates session summaries for context bridging

Start the Daemon

Run in background using Task tool:

Task tool:
  subagent_type: "general-purpose"
  run_in_background: true
  description: "Kanban review loop daemon"
  prompt: [See daemon instructions below]

Daemon Instructions

You are the KANBAN REVIEW LOOP DAEMON with Ralph Wiggum awareness and session tracking.

## Main Loop (repeat continuously)

### Step 0: Session Check (once at start)
Start session: kanban_session_start with agentId: "review-loop-daemon"
This gives you context about recent activity and any continuity notes.

### Step 1: Board Health Verification
Run: kanban_verify_board_health
This returns structured health status with recommendation.

### Step 2: Full Health Check
Run: kanban_health_check with role: "architect", staleThresholdHours: 24

### Step 3: Check Escalations
Run: kanban_get_escalated_tasks with role: "architect"

### Step 4: Check Active Sprint
Run: kanban_sprint_list with role: "architect"
For each active sprint, check iteration count vs maxIterations.

### Step 5: Analyze & Alert

Check for issues:

**CRITICAL ALERTS:**
- **Escalated tasks** (exceeded maxIterations):
  Log "CRITICAL: Task [title] escalated after [N] iterations - needs human review"
- **Critical tasks not started**:
  Log "ALERT: Critical task [title] waiting"

**WARNING ALERTS:**
- **High iteration tasks** (iteration >= maxIterations - 1):
  Log "WARNING: Task [title] at iteration [N]/[max] - at risk of escalation"
- **Stale tasks** (in_progress > 24h):
  Log "WARNING: Task [title] stale for [hours]h"
- **Sprint at risk** (iteration near max):
  Log "WARNING: Sprint [goal] at iteration [N]/[max]"

**INFO ALERTS:**
- **QA backlog > 3**:
  Log "INFO: QA backlog high, spawning QA agent"
- **Overloaded agents** (> 5 tasks):
  Log "INFO: Agent [id] overloaded with [N] tasks"
- **Low backlog** (< 3 tasks):
  Log "INFO: Backlog low, plan more work"

### Step 6: Auto-Remediation

If QA backlog > 3, spawn QA agent (NOT in background):

Task tool: subagent_type: "general-purpose" prompt: | You are QA. Clear the backlog. 1. kanban_session_start with agentId: "qa-auto" 2. kanban_qa_list with role: "qa" 3. For each task: - kanban_get_task_detail to see iteration history - Review and approve/reject with structured feedback 4. kanban_session_end with summary


### Step 7: Generate Summary
Run: kanban_generate_summary
This updates the session-summary.md file for other agents.

### Step 8: Report Status

Output:
---
[TIME] Review Loop Check #N
Health: [OK | WARNING | CRITICAL]

Sprint Status:
- [Sprint goal]: Iteration [N]/[max] - [status]

Board:
- Backlog: X | In Progress: Y | Blocked: Z | Done: W
- Pending QA: Q
- Escalated: E

Issues:
- [list of alerts]

Actions Taken:
- [any remediation]

Next check in 5 minutes...
---

### Step 9: Sleep
Wait 5 minutes, then repeat from Step 1.

### On Shutdown
When loop is stopped:

kanban_session_end with: agentId: "review-loop-daemon" sessionNotes: "Review loop stopped. X checks performed." cleanState: true

Monitoring

After starting:

  1. Note the task ID returned
  2. Check status: TaskOutput with task_id, block: false
  3. Stop: User kills the background task

Alert Priority

Level Action
CRITICAL Immediate notification, needs human intervention
WARNING Soon to become critical, monitor closely
INFO Informational, auto-remediation if possible

Examples

User: "/kanban-review-loop"
-> Spawn background daemon
-> Report task ID for monitoring
-> Daemon runs continuously:
   -> kanban_session_start (once)
   -> Checks board health with kanban_verify_board_health
   -> Monitors escalations
   -> Tracks sprint iterations
   -> Alerts on issues
   -> Auto-spawns QA when needed (with session protocols)
   -> kanban_generate_summary (updates file)
   -> Sleep 5 minutes, repeat
-> On stop: kanban_session_end

Didn't find tool you were looking for?

Be as detailed as possible for better results