deploy
Get the feature deployed to staging (preferred) or production using the project's real mechanism, with preflight, confirmation, and verification.
Skill by jshepelsky
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/jshepelsky-deploy
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: devops
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/deploy
What's inside
- Skill name
deploy- Description
- Get the feature deployed to staging (preferred) or production using the project's real mechanism, with preflight, confirmation, and verification. Step 15 of /ship.
- License
- MIT
- Instructions
- 62 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
Deploy: $ARGUMENTS `$ARGUMENTS` names the target environment. Default to staging when the profile has one; use production only when it's the only environment or the user names it. Hard rules: - **Never commit, branch, push, tag, or merge.** If the deploy mechanism is triggered by git (push to a branch, a tag, a merged PR), you prepare everything, tell the user the exact git action to take, and then watch the result. You don't take the git action. - **Never deploy to production without the user saying yes to the exact command in this conversation.** No flags override this. - Never deploy a dirty working tree (uncommitted source changes), a failing suite, or code the user hasn't committed. - Never run destructive data operations as part of a deploy without a separate confirmation. Read the profile's `## Deploy` section. If it's missing, do Step 1, then offer to run `/first-run` to record it. ## Step 1 - Detect the deploy mechanism (only if the profile lacks it) ```bash ls .github/workflows .gitlab-ci.yml .circleci bitbucket-pipelines.yml 2>/dev/null ls vercel.json .vercel netlify.toml fly.toml Procfile app.json render.yaml railway.json serverless.yml wrangler.toml firebase.json amplify.yml 2>/dev/null ls Dockerfile docker-compose*.yml Capfile deploy.rb k8s kubernetes helm charts terraform 2>/dev/null grep -n -i -E '"deploy|release|publish' package.json 2>/dev/null; grep -n -i -E '^(deploy|release|publish)' Makefile justfile 2>/dev/null ls bin/deploy* scripts/deploy* deploy.sh 2>/dev/null grep -n -i -A5 'deploy' README* CONTRIBUTING* docs/*.md 2>/dev/null | head -60 ``` For CI-driven deploys, open the workflow and find the trigger. Record per environment: trigger or command, URL, health-check URL, logs command, pre-deploy steps (migrations, build, cache), where secrets live, rollback method. Mark guesses `(unconfirmed)`. ## Step 2 - Preflight Print a checklist; stop at the first failure. 1. **Tree.** `git status --porcelain` shows no source changes (the work folder is fine). If there are uncommitted source changes, stop: the user commits first. 2. **Branch.** You're on the branch the environment deploys from, or the user has told you which branch to deploy. `git fetch`; report if local is ahead of or behind the remote. If ahead, the user needs to push. 3. **Suite.** `/test-all --final` has been ticked in `status.md` for this exact tree. If not, run it now. 4. **Migrations.** List pending migrations. For each: reversible, locks, works with both old and new code. If migrations run separately from deploy, note the step. 5. **Config.** Diff env vars and config keys referenced by the feature against the last deployed commit. Check they exist in the target env if you can (`vercel env ls`, `fly secrets list`, `heroku config`); otherwise list them for the user to confirm. 6. **Release bookkeeping.** If the project keeps a changelog or version, run `/changelog` and prepare the version bump as an uncommitted change for the user to commit. 7. **What ships.** `git log --oneline <last-deployed>..HEAD`. ## Step 3 - Deploy - **Command-driven** (`vercel`, `fly deploy`, `cap`, `make deploy`, a script): show the exact command; for production, wait for a yes. Run it and capture output. - **Git-driven** (push, tag, merge): tell the user exactly what to do (`git push origin main`, `git tag v1.4.0 && git push --tags`, "merge PR #12"). Wait for them to say it's done, then watch the pipeline: ```bash gh run list --limit 1 gh run watch <id> --exit-status ``` - Run migrations if they're a separate step, in the profile's order. - If the deploy fails, stop and report the output. Don't retry blindly. ## Step 4 - Verify 1. Health check returns 2xx. The version or commit endpoint, if any, reports the expected SHA. 2. Load the feature's new screens and routes once. 3. Tail logs for about a minute. If verification fails: say so first, show evidence, propose the profile's rollback, and roll back only on a yes. ## Step 5 - Report Environment, SHA, URL, what was verified, migrations run, the rollback command ready to paste. Tick checkpoint 15 in `status.md`. Next: `/env-test <env>`.
From jshepelsky/claude-essential-plugin, MIT.
More from jshepelsky/claude-essential-plugin
- 0.565score
flaky-test-investigation
jshepelsky
Investigate a flaky or failing test using evidence before proposing a fix.
SkillAFreshness: AClaude Code +9 - 0.565score
humanize
jshepelsky
Rewrite text to remove signs of AI-generated writing — inflated significance, promotional language, em-dash overuse, rule of three, AI vocabulary, copula…
SkillAFreshness: AClaude Code +9 - 0.561score
env-test
jshepelsky
Run the feature's tests against a deployed environment - full suites on staging, read-only and data-safe checks on production.
SkillAFreshness: AClaude Code +9 - 0.561score
phase-plan
jshepelsky
Order the work into phases with dependencies and blockers so development runs in the most efficient sequence.
SkillAFreshness: AClaude Code +9 - 0.561score
develop
jshepelsky
Build one phase from the phase plan - code, planned tests, every test type, per-phase code review - then hand off with a branch name, commit message, and a…
SkillAFreshness: AClaude Code +9 - 0.561score
document
jshepelsky
Write and audit documentation after a feature ships - CLAUDE.md, process docs, and anything a future session or developer needs.
SkillAFreshness: AClaude Code +9 - 0.561score
mockup
jshepelsky
Create HTML mockups when no design was supplied, with real research, then review them with a critical eye.
SkillAFreshness: AClaude Code +9 - 0.561score
dev-setup
jshepelsky
Create and refresh the files that help Claude develop this feature well - CLAUDE.md, the essentials profile, and per-area notes.
SkillAFreshness: AClaude Code +9 - 0.558score
commit
jshepelsky
Stage logical change groups and write a clean commit message from the diff.
SkillAFreshness: AClaude Code +9 - 0.558score
first-run
jshepelsky
One-time setup — detect this codebase and write a profile the essentials agents specialize to.
SkillAFreshness: AClaude Code +9 - 0.558score
changelog
jshepelsky
Generate release notes from git history since the last tag (or a given range).
SkillAFreshness: AClaude Code +9 - 0.553score
dependency-audit
jshepelsky
Audit dependencies for known vulnerabilities, outdated versions, and unused packages.
SkillAFreshness: AClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Oct 22, 2026 (7 days ago)
- Last release
- Oct 22, 2026 (7 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from jshepelsky
- 0.565score
flaky-test-investigation
jshepelsky
Investigate a flaky or failing test using evidence before proposing a fix.
SkillAFreshness: AClaude Code +9 - 0.565score
humanize
jshepelsky
Rewrite text to remove signs of AI-generated writing — inflated significance, promotional language, em-dash overuse, rule of three, AI vocabulary, copula…
SkillAFreshness: AClaude Code +9 - 0.561score
dev-setup
jshepelsky
Create and refresh the files that help Claude develop this feature well - CLAUDE.md, the essentials profile, and per-area notes.
SkillAFreshness: AClaude Code +9 - 0.561score
env-test
jshepelsky
Run the feature's tests against a deployed environment - full suites on staging, read-only and data-safe checks on production.
SkillAFreshness: AClaude Code +9 - 0.561score
phase-plan
jshepelsky
Order the work into phases with dependencies and blockers so development runs in the most efficient sequence.
SkillAFreshness: AClaude Code +9 - 0.561score
develop
jshepelsky
Build one phase from the phase plan - code, planned tests, every test type, per-phase code review - then hand off with a branch name, commit message, and a…
SkillAFreshness: AClaude Code +9
Cite this page
deploy skill. Agent Mart. https://agentmart.dev/skills/jshepelsky-deploy (last updated 2026-10-24)