Semantic Release

semantic-release

Overview

Automate semantic versioning and release management using semantic-release v25+ (Node.js) following 2025 best practices. Works with all languages (JavaScript, TypeScript, Python, Rust, Go, C++, etc.) via the @semantic-release/exec plugin. Create shareable configurations for multi-repository setups, initialize individual projects with automated releases, and configure GitHub Actions workflows with OIDC trusted publishing.

Important: This skill uses semantic-release (Node.js) exclusively, NOT python-semantic-release, even for Python projects. Rationale: 23.5x larger community, 100x+ adoption, better future-proofing.

When to Use This Skill

Invoke when:

• Setting up local releases for a new project (any language)
• Creating shareable semantic-release configuration for organization-wide use
• Migrating existing projects to 2025 semantic-release patterns
• Troubleshooting semantic-release setup or version bumps
• Setting up Python projects (use Node.js semantic-release, NOT python-semantic-release)
• Configuring GitHub Actions (optional backup, not recommended as primary due to speed)
• Rust workspaces using release-plz (see Rust reference)

Why Node.js semantic-release

22,900 GitHub stars - Large, active community 1.9M weekly downloads - Proven adoption 126,000 projects using it - Battle-tested at scale 35+ official plugins - Rich ecosystem Multi-language support - Works with any language via @semantic-release/exec

Do NOT use python-semantic-release. It has a 23.5x smaller community (975 vs 22,900 stars), ~100x less adoption, and is not affiliated with the semantic-release organization.

Release Workflow Philosophy: Local-First

Default approach: Run releases locally, not via GitHub Actions.

Why Local Releases

Primary argument: GitHub Actions is slow

• ⏱️ GitHub Actions: 2-5 minute wait for release to complete
• ⚡ Local release: Instant feedback and file updates
• 🔄 Immediate workflow continuity - no waiting for CI/CD

Additional benefits:

• ✅ Instant local file sync - package.json, CHANGELOG.md, tags updated immediately
• ✅ No pull required - Continue working without git pull after release
• ✅ Dry-run testing - npm run release:dry to preview changes before release
• ✅ Offline capable - Can release without CI/CD dependency
• ✅ Faster iteration - Debug release issues immediately, not through CI logs

GitHub Actions: Optional Backup Only

GitHub Actions workflows are provided as optional automation, not the primary method:

• Use for team consistency if required
• Backup if local environment unavailable
• Not recommended as primary workflow due to speed

Authentication Setup

gh auth login
# Browser authentication once
# Credentials stored in keyring
# All future releases: zero manual intervention

This is the minimum manual intervention possible for local semantic-release with GitHub plugin functionality.

Multi-Account Authentication via mise [env]

For multi-account GitHub setups, use mise [env] to set per-directory GH_TOKEN:

# ~/your-project/.mise.toml
[env]
GH_TOKEN = "{{ read_file(path=env.HOME ~ '/.claude/.secrets/gh-token-accountname') | trim }}"
GITHUB_TOKEN = "{{ read_file(path=env.HOME ~ '/.claude/.secrets/gh-token-accountname') | trim }}"

This overrides gh CLI's global authentication, ensuring semantic-release uses the correct account for each directory.

See the mise-configuration skill for complete setup.

mise Task Detection

When .mise.toml has release tasks, prefer mise run over npm run:

See Python Guide for complete mise workflow example.

GitHub Actions Policy

CRITICAL: No testing or linting in GitHub Actions. See CLAUDE.md for full policy.

Separation of Concerns (4-Level Architecture)

semantic-release configuration follows a hierarchical, composable pattern:

Level 1: Skill - ${CLAUDE_PLUGIN_ROOT}/skills/semantic-release/ (Generic templates, system-wide tool) Level 2: User Config - ~/semantic-release-config/ (@username/semantic-release-config) Level 3: Organization Config - npm registry (@company/semantic-release-config) Level 4: Project Config - .releaserc.yml in project root

Configuration Precedence

Level 4 (Project) → overrides → Level 3 (Org) → overrides → Level 2 (User) → overrides → Defaults

Conventional Commits Format

semantic-release analyzes commit messages to determine version bumps:

<type>(<scope>): <subject>

Version Bump Rules (Default)

• feat: → MINOR version bump (0.1.0 → 0.2.0)
• fix: → PATCH version bump (0.1.0 → 0.1.1)
• BREAKING CHANGE: or feat!: → MAJOR version bump (0.1.0 → 1.0.0)
• docs:, chore:, style:, refactor:, perf:, test: → No version bump (by default)

Release Notes Visibility (Important)

Warning: The @semantic-release/release-notes-generator (Angular preset) only includes these types in release notes:

• feat: → Features section
• fix: → Bug Fixes section
• perf: → Performance Improvements section

Other types (docs:, chore:, refactor:, etc.) trigger releases when configured but do NOT appear in release notes.

Recommendation: For documentation changes that should be visible in release notes, use:

fix(docs): description of documentation improvement

This ensures the commit appears in the "Bug Fixes" section while still being semantically accurate (fixing documentation gaps is a fix).

Marketplace Plugin Configuration (Always Bump)

For Claude Code marketplace plugins, every change requires a version bump for users to receive updates.

Option A: Shareable Config (if published)

# .releaserc.yml
extends: "@terryli/semantic-release-config/marketplace"

Option B: Inline Configuration

# .releaserc.yml
plugins:
  - - "@semantic-release/commit-analyzer"
    - releaseRules:
        # Marketplace plugins require version bump for ANY change
        - { type: "docs", release: "patch" }
        - { type: "chore", release: "patch" }
        - { type: "style", release: "patch" }
        - { type: "refactor", release: "patch" }
        - { type: "test", release: "patch" }
        - { type: "build", release: "patch" }
        - { type: "ci", release: "patch" }

Result after configuration:

Why marketplace plugins need this: Plugin updates are distributed via version tags. Without a version bump, users running /plugin update see no changes even if content was modified.

MANDATORY: Every Release Must Increment Version

Pre-release validation: Before running semantic-release, verify releasable commits exist since last tag. A release without version increment is invalid.

Autonomous check sequence:

1. List commits since last tag: compare HEAD against latest version tag
2. Identify commit types: scan for feat:, fix:, or BREAKING CHANGE: prefixes
3. If NO releasable commits found → STOP - do not proceed with release
4. Inform user: "No version-bumping commits since last release. Use feat: or fix: prefix for releasable changes."

Commit type selection guidance:

• Use fix: for any change that improves existing behavior (bug fixes, enhancements, documentation corrections that affect usage)
• Use feat: for new capabilities or significant additions
• Reserve chore:, docs:, refactor: for changes that truly don't warrant a release

Why this matters: A release without version increment creates confusion - users cannot distinguish between releases, package managers may cache old versions, and changelog entries become meaningless.

MAJOR Version Breaking Change Confirmation

Trigger: BREAKING CHANGE: footer or feat!: / fix!: prefix in commits.

When MAJOR is detected, this skill runs a 3-phase confirmation workflow:

1. Detection: Scan commits for breaking change markers
2. Analysis: Spawn 3 parallel subagents (User Impact, API Compat, Migration)
3. Confirmation: AskUserQuestion with proceed/downgrade/abort options

See MAJOR Confirmation Workflow for complete details including subagent prompts, decision tree, and example output.

Examples

Feature (MINOR):

feat: add BigQuery data source support

Bug Fix (PATCH):

fix: correct timestamp parsing for UTC offsets

Breaking Change (MAJOR):

feat!: change API to require authentication

BREAKING CHANGE: All API calls now require API key in Authorization header.

Documentation Linking

Auto-include doc changes in release notes. Add to .releaserc.yml:

- - "@semantic-release/exec"
  - generateNotesCmd: "node plugins/itp/skills/semantic-release/scripts/generate-doc-notes.mjs ${lastRelease.gitTag}"

Detects: ADRs, Design Specs, Skills, Plugin READMEs. See Doc Release Linking.

Note: The @semantic-release/exec plugin uses Lodash templates (${var}). This conflicts with bash default syntax (${VAR:-default}). See Troubleshooting: Lodash Template Conflicts.

Quick Start

Prerequisites

Initialize Project

./scripts/init-project.mjs --project   # Initialize current project
./scripts/init-project.mjs --user      # Create user-level shareable config
./scripts/init-project.mjs --help      # See all options

Run Release

See Local Release Workflow for the complete 4-phase process.

Python Projects

semantic-release handles versioning. For PyPI publishing, see pypi-doppler skill.

Version pattern (importlib.metadata - never hardcode):

from importlib.metadata import PackageNotFoundError, version
try:
    __version__ = version("your-package-name")
except PackageNotFoundError:
    __version__ = "0.0.0+dev"

See Python Projects Guide for complete setup including Rust+Python hybrids.

GitHub Actions (Optional)

Not recommended as primary (2-5 minute delay). Repository Settings → Actions → Workflow permissions → Enable "Read and write permissions".

Reference Documentation

Cross-skill references:

• mise-tasks skill: polyglot-affected - Complete Pants + mise integration guide
• mise-tasks skill: bootstrap-monorepo - Autonomous polyglot monorepo setup
• pypi-doppler skill - Local PyPI publishing with Doppler

Post-Change Checklist

After modifying THIS skill (semantic-release):

1. SKILL.md and references remain aligned
2. New references documented in Reference Documentation table
3. All referenced files in references/ exist
4. Append changes to evolution-log.md
5. Validate with bun scripts/validate-plugins.mjs
6. Run npm run release:dry to verify no regressions