AgentSkillsCN

remediation

Kubernetes 安全修复措施。在提出或执行 Pod 重启、部署扩缩容，或回滚操作时使用。务必先进行模拟演练，确保万无一失。

SKILL.md
--- frontmatter
name: remediation
description: Safe remediation actions for Kubernetes. Use when proposing or executing pod restarts, deployment scaling, or rollbacks. Always use dry-run first.

Remediation Actions

Safety Principles

  1. •ALWAYS dry-run first - All scripts support --dry-run flag
  2. •Confirm before executing - Show what will happen, ask for confirmation
  3. •Document the action - Log what was done and why
  4. •Have a rollback plan - Know how to undo the action

Available Scripts

All scripts are in .claude/skills/remediation/scripts/

restart_pod.py - Restart a pod by deleting it

bash
# Dry run (shows what would happen)
python .claude/skills/remediation/scripts/restart_pod.py <pod-name> -n <namespace> --dry-run

# Execute
python .claude/skills/remediation/scripts/restart_pod.py <pod-name> -n <namespace>

scale_deployment.py - Scale a deployment

bash
# Dry run
python .claude/skills/remediation/scripts/scale_deployment.py <deployment> -n <namespace> --replicas N --dry-run

# Execute
python .claude/skills/remediation/scripts/scale_deployment.py <deployment> -n <namespace> --replicas N

rollback_deployment.py - Rollback to previous revision

bash
# Dry run (shows current and target revision)
python .claude/skills/remediation/scripts/rollback_deployment.py <deployment> -n <namespace> --dry-run

# Execute
python .claude/skills/remediation/scripts/rollback_deployment.py <deployment> -n <namespace>

Remediation Workflow

  1. •Diagnose first - Use k8s-debugger to understand the issue
  2. •Propose action - State what you plan to do and why
  3. •Dry run - Show what will happen
  4. •Get confirmation - Ask user to confirm
  5. •Execute - Run the action
  6. •Verify - Check that the issue is resolved

Common Remediation Scenarios

Pod stuck in CrashLoopBackOff

bash
# 1. Check events
python .claude/skills/infrastructure/kubernetes/scripts/get_events.py <pod> -n <namespace>

# 2. If fixable by restart, dry-run first
python .claude/skills/remediation/scripts/restart_pod.py <pod> -n <namespace> --dry-run

# 3. Execute restart
python .claude/skills/remediation/scripts/restart_pod.py <pod> -n <namespace>

Deployment stuck with bad image

bash
# 1. Check history
python .claude/skills/infrastructure/kubernetes/scripts/get_history.py <deployment> -n <namespace>

# 2. Dry-run rollback
python .claude/skills/remediation/scripts/rollback_deployment.py <deployment> -n <namespace> --dry-run

# 3. Execute rollback
python .claude/skills/remediation/scripts/rollback_deployment.py <deployment> -n <namespace>

Service under high load

bash
# 1. Check current state
python .claude/skills/infrastructure/kubernetes/scripts/describe_deployment.py <deployment> -n <namespace>

# 2. Dry-run scale up
python .claude/skills/remediation/scripts/scale_deployment.py <deployment> -n <namespace> --replicas 5 --dry-run

# 3. Execute scale
python .claude/skills/remediation/scripts/scale_deployment.py <deployment> -n <namespace> --replicas 5

Output Format

When proposing remediation, use this structure:

code
## Proposed Remediation

**Action**: [e.g., Restart pod, Scale deployment, Rollback]
**Target**: [resource name and namespace]
**Reason**: [why this action will help]
**Risk**: [potential side effects]

### Dry Run Output
[output from --dry-run]

### Confirmation Required
Please confirm you want to proceed with this action.