Deploy Pr

Deploy PR

Deploys a merged pull request to the target environment, runs smoke tests to verify the deployment, and confirms a rollback plan exists. Bridges the gap between merging code and having it running in production.

Prerequisites

• Apply the shared SDLC conventions in skills/sdlc/references/shared.md.
• If no argument is provided, target the pull request from $PR_NUMBER (and $REPO).
• gh CLI authenticated with write access to the target repository
• A merged PR number or merge SHA provided as $1
• Deployment tooling available (CI/CD pipeline, container registry, or manual deployment scripts)
• Read any files present under .sdlc/context/ for project-level context, especially architecture.md for deployment topology

Workflow

Fetch merged PR details ($1)
        |
        v
  Identify target environment
        |
        v
  Confirm rollback plan exists
     /         \
   Yes          No
    |            |
    v            v
  Execute     Document rollback
  deployment  procedure first,
    |         then proceed
    v
  Run smoke tests
     /         \
   Pass         Fail
    |            |
    v            v
  Verify      Execute rollback
  health      Report failure
    |            |
    v            v
  Report      Stop, investigate
  success

Steps

1. Fetch the merged PR details:

`` gh pr view $1 --json title,body,mergeCommit,headRefName,closingIssuesReferences ``

1. Identify the target environment(s) from the PR labels, branch name, or project configuration. Common targets:

• staging for non-production validation
• production for user-facing releases
• Feature-specific environments (e.g., preview, canary)

If uncertain, ask the user which environment to deploy to.

1. Check for a rollback plan. Look for:

• A documented rollback procedure in the PR description or .sdlc/ artifacts
• Previous deployment version or tag to roll back to
• Database migration rollback if applicable

If no rollback plan exists, document one before proceeding.

1. Execute the deployment using the project's CI/CD pipeline or deployment scripts:

`` gh workflow run deploy.yml --ref main -f environment= ` Or, if deployment is manual, follow the project's deployment procedure from architecture.md`.

1. Wait for deployment to complete. Monitor:

`` gh run watch ``

1. Run smoke tests against the deployed environment:

• Health check endpoints return 200
• Key user flows are accessible
• No increase in error rates
• Database connectivity is intact

Use the project's existing smoke test suite if available, otherwise test the critical path manually.

1. If smoke tests pass, verify overall system health:

• Check monitoring dashboards for anomalies
• Confirm no new alerts fired
• Verify key metrics (latency, error rate, throughput) are within normal ranges

1. If smoke tests fail, execute the rollback plan:

• Redeploy the previous version
• Run any database rollback migrations
• Confirm the system is healthy on the previous version
• Report the failure with details for investigation

1. Report deployment status:

• Environment deployed to
• Version or commit SHA deployed
• Smoke test results
• Rollback plan location
• Any issues encountered

Output Format

## Deployment Report

- **PR:** # — 
- **Commit:** 
- **Environment:** 
- **Deployed at:** 
- **Smoke tests:** Pass / Fail ()
- **Rollback plan:** 
- **Status:** Success / Rolled back / Failed

Example Usage

Scenario 1: Standard production deploy

/deploy-pr 42

PR #42 was merged to main. Deploy to production via CI/CD pipeline, run smoke tests, confirm healthy.

Scenario 2: Staging deploy first

/deploy-pr 88

Deploy PR #88 to staging first, run smoke tests, then promote to production if all checks pass.

Scenario 3: Rollback needed

/deploy-pr 55

Smoke tests reveal a 500 error on the checkout flow. Roll back to the previous deployment, confirm system health, report the failure.

Next Step

Once deployment is successful, continue with /create-learnings to capture a retrospective on the feature. If the feature is now running in production, consider running /observe-production to establish a baseline for monitoring.

Useful Commands Reference

| Command | Description | |---|---| | gh pr view --json title,body,mergeCommit | Fetch merged PR details | | gh workflow run deploy.yml --ref main -f environment= | Trigger deployment workflow | | gh run watch | Watch CI/CD run until completion | | gh run list --workflow=deploy.yml --limit=5 | List recent deployment runs |

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: tomzx
• Source: tomzx/agents
• License: MIT

Install and usage instructions live in the source repository linked above.