cicd-pipeline-monitor

name	cicd-pipeline-monitor
description	Use this skill to monitor, trigger, diagnose, and remediate CI/CD pipelines across GitHub Actions, Azure DevOps, Jenkins, and ArgoCD. Triggers: any request to check pipeline status, investigate a build failure, re-trigger a deployment, analyse flaky tests, enforce pipeline standards, generate a deployment summary report, or show DORA metrics, deployment frequency, change failure rate, mean time to restore.
tools	["bash","computer"]

CI/CD Pipeline Monitor Skill

Provide full observability and automated remediation for CI/CD pipelines. Connect to pipeline APIs, parse run results, surface root-cause analysis, and take corrective action within defined safe-action boundaries.

---

Supported Platforms

Platform	Auth method	API base
GitHub Actions	GITHUB_TOKEN env var	https://api.github.com
Azure DevOps	ADO_PAT env var	https://dev.azure.com/{org}
Jenkins	JENKINS_USER/TOKEN	http://{host}/api/json
ArgoCD	ARGOCD_TOKEN	https://{host}/api/v1

---

Core Workflows

Check Pipeline Status

# GitHub Actions
gh run list --repo $REPO --limit 10 --json status,conclusion,name,createdAt

# Azure DevOps
az pipelines runs list --org $ADO_ORG --project $PROJECT --top 10

# ArgoCD
argocd app list --output json

Parse and return:

• Run ID, pipeline name, trigger, status, duration
• Last success timestamp
• Consecutive failure count

Investigate a Failed Run

1. Fetch the failed run's logs
2. Extract the failing step and error block (last 50 lines of failing job)
3. Classify error type:
   • Flaky test → flag for retry, log to flake tracker
   • Infra timeout → retry once, then escalate
   • Config/code error → open issue with full context
   • Dependency failure → check upstream pipeline / registry health
4. Return structured diagnosis

Trigger / Re-trigger

# GitHub Actions
gh workflow run $WORKFLOW --repo $REPO --ref $BRANCH

# Azure DevOps
az pipelines run --id $PIPELINE_ID --branch $BRANCH

# ArgoCD sync
argocd app sync $APP_NAME --force

Only auto-trigger re-runs for classified flaky/infra failures. Code/config failures require human review first.

Deployment Summary Report

Aggregate across all pipelines for a time window:

Deployments:     total | successful | failed | rolled-back
Mean lead time:  X minutes
Change fail rate: X%
MTTR:            X minutes
Top failing pipelines: [list]

---

Alerting Thresholds

Metric	Warning	Critical
Consecutive failures	2	3
Pipeline duration (P95)	+50%	+100%
Change fail rate (7-day)	>10%	>25%
Queue depth	>20	>50

On threshold breach: emit a structured alert → route to incident channel.

---

Pipeline Standards Enforcement

When reviewing or generating pipelines, enforce:

• All secrets via vault/secret-store — never plaintext in YAML
• Branch protection: main/prod require passing checks before merge
• Docker images pinned to digest, not latest
• Scan steps: SAST (e.g., semgrep), container scan (trivy), IaC scan (checkov)
• Separate build / test / deploy stages with explicit approval gates for prod
• Artifacts retained for minimum 30 days

---

Safe Action Boundaries

Action	Auto	Requires Approval
Re-run flaky test	✅
Re-run infra timeout	✅
Cancel stuck run (>2h)	✅
Trigger prod deployment		✅
Modify pipeline YAML		✅
Disable a pipeline		✅

---

Examples

• "What failed in the last prod deployment?"
• "Show me all pipelines that failed more than twice this week"
• "Re-run the flaky integration tests on the payments service"
• "Generate a DORA metrics report for the last 30 days"
• "Check if the ArgoCD sync for tenant-42 completed successfully"

---

Output Format

{
  "pipeline_tool": "github-actions|azuredevops|jenkins|argocd",
  "pipeline": "string",
  "failure_reason": "string",
  "status": "success|failure|running|queued",
  "last_run": "ISO8601",
  "diagnosis": { "error_type": "string", "failing_step": "string", "message": "string" },
  "action_taken": "none|retried|escalated",
  "metrics": { "lead_time_min": 0, "fail_rate_pct": 0 },
  "dora": { "deployment_frequency": 0, "lead_time": 0, "mttr": 0, "change_fail_rate": 0 }
}