AgentSkillsCN

deploy

为某项功能触发部署工作流。调用 SRE 代理，负责部署配置、预发布阶段的灰度 rollout，以及生产环境的正式上线。

SKILL.md
--- frontmatter
name: deploy
description: >
  Trigger deployment workflow for a feature. Invokes the SRE agent to handle
  deployment configuration, staging rollout, and production promotion.
tools: Read, Grep, Glob, Bash
model: sonnet
disable-model-invocation: true
context: fork
agent: general-purpose
argument-hint: "<feature-id> [--promote|--hotfix]"

Deploy Command

Orchestrate deployment through the pipeline with proper gates and validation.

Usage

code
/deploy <feature-id>                  Deploy feature to staging
/deploy <feature-id> --promote        Promote from staging to production
/deploy <feature-id> --hotfix         Abbreviated pipeline for urgent fixes

Arguments

Feature ID: $ARGUMENTS

Prerequisites

Before deployment:

  • •Review gate must be approved: /pipeline approve <id> review
  • •All blocking findings must be resolved
  • •Tests must pass

Workflow

/deploy <feature-id>

  1. •

    Validate prerequisites

    • •Check review gate is approved
    • •Verify no blocking findings
    • •Confirm tests pass (check CI status)
  2. •

    Load deployment context

    • •Read design artifact for infrastructure requirements
    • •Read service profile for deployment targets
    • •Check FluxCD configuration requirements
  3. •

    Invoke SRE agent

    • •Generate/update Kustomize overlays
    • •Configure staging deployment
    • •Set up monitoring and alerts
  4. •

    Update pipeline state

    • •Mark deploy stage as in_progress
    • •Record deployment timestamp

/deploy <feature-id> --promote

  1. •

    Validate staging success

    • •Check staging deployment is healthy
    • •Verify staging tests pass
    • •Confirm metrics are nominal
  2. •

    Invoke SRE agent for production

    • •Update production overlay
    • •Configure canary/rollout strategy
    • •Set up production alerts
  3. •

    Trigger document and announce stages

    • •These can run in parallel with production monitoring

/deploy <feature-id> --hotfix

Abbreviated path for urgent fixes:

  1. •Skip normal gates (with audit trail)
  2. •Direct to production with canary rollout
  3. •Require post-deploy review within 24 hours
  4. •Auto-create follow-up tasks for documentation

Deployment Context Template

markdown
## Deployment Context

**Feature**: {id} - {name}
**Target**: {staging|production}
**Review Status**: Approved on {date}

### Infrastructure Requirements

{From design artifact}

### Deployment Configuration

{From service profile and kustomize patterns}

### Rollout Strategy

- Canary: {percentage}
- Rollout duration: {time}
- Rollback trigger: {conditions}

### Monitoring

- Key metrics to watch
- Alert thresholds
- Dashboard links

Output

code
Deployment initiated: feat-042-vm-snapshot-management
Target: staging
Review gate: Approved (2025-01-15)

[Invokes SRE agent]

Deployment configured.
Kustomize overlay: config/overlays/staging/feat-042/
FluxCD will reconcile within 5 minutes.

Monitor: /deploy status feat-042
Promote to production: /deploy feat-042 --promote

Error Handling

Review not approved:

code
Cannot deploy feat-042: review gate not approved.
Action: Run /pipeline approve feat-042 review after resolving findings.

Blocking findings exist:

code
Cannot deploy feat-042: 2 blocking findings unresolved.
Findings:
  - missing-status-condition in pkg/apis/snapshots/v1alpha1/types.go
  - unvalidated-input in pkg/registry/snapshot/strategy.go
Action: Resolve blocking findings and request re-review.

Tests failing:

code
Cannot deploy feat-042: CI tests failing.
Failed jobs: unit-tests, integration-tests
Action: Fix failing tests before deployment.