git-conventional-commits（git 常规提交）

Git Conventional Commits

Overview

Generate structured commit messages following the Conventional Commits specification. The format is type(scope): description with an optional body and footer. This skill ensures agents produce consistent, parseable commit messages that work with semantic versioning tools, changelogs, and CI pipelines.

When to use

• The task is to write a conventional commit message for currently staged git changes.
• The user wants help classifying a change as feat, fix, refactor, docs, and so on.
• The deliverable is a commit subject/body/footer, not a PR description or changelog.
• The repository expects or appears compatible with Conventional Commits.

Do NOT use when:

• There are no staged changes to classify.
• The user already supplied the exact commit message they want to use.
• The repository follows a different commit convention that should be preserved.

Response format

Always structure the final response with these top-level sections, in this order:

1. Summary — state the task, scope, and main conclusion in 1-3 sentences.
2. Decision / Approach — state the key classification, assumptions, or chosen path.
3. Artifacts — provide the primary deliverable(s) for this skill. Use clear subheadings for multiple files, commands, JSON payloads, queries, or documents.
4. Validation — state checks performed, important risks, caveats, or unresolved questions.
5. Next steps — list concrete follow-up actions, or write None if nothing remains.

Rules:

• Do not omit a section; write None when a section does not apply.
• If files are produced, list each file path under Artifacts before its contents.
• If commands, JSON, SQL, YAML, or code are produced, put each artifact in fenced code blocks with the correct language tag when possible.
• Keep section names exactly as written above so output stays predictable across skills.

Workflow

1. Read the staged diff

Run git diff --cached to see exactly what will be committed. If nothing is staged, tell the user and stop.

Also run git diff --cached --stat for a file-level summary. This helps identify scope.

Why this matters: The diff is the ground truth. Never guess what changed — read it.

2. Classify the change type

Analyze the diff and select ONE primary type. Use the first matching rule:

Decision rules for ambiguous cases:

• New file with a function → feat (new capability), not chore
• Bug fix that also refactors → fix (the primary intent matters)
• Test added alongside a feature → feat (the feature is the main change)
• Dependency update to fix a vulnerability → fix, not build
• Renaming a public API method → refactor with BREAKING CHANGE footer

3. Identify the scope

The scope is the module, component, or area affected. It goes in parentheses: type(scope):...

How to determine scope:

• If all changes are in one directory/module → use that name (e.g., auth, api, parser)
• If changes span a single feature across files → use the feature name (e.g., login, search)
• If changes are too broad to name → omit the scope: type: description

Scope should be a single lowercase word or hyphenated phrase. Never use file paths as scope.

4. Write the subject line

The subject line is the most important part. Follow these rules strictly:

1. Use imperative mood — "add feature" not "added feature" or "adds feature". Write as if completing the sentence: "If applied, this commit will ___."
2. 50 characters or fewer — Hard limit. Count the ENTIRE first line including type(scope):. For example, feat(auth): add login endpoint is 31 characters. If you're over 50, shorten the description or move details to the body.
3. No period at the end — It wastes a character and is unconventional.
4. Lowercase first letter after the colon — feat(auth): add login endpoint not feat(auth): Add login endpoint.
5. Be specific — "fix null pointer in user lookup" not "fix bug".
6. Verify before outputting — Count the characters of your subject line. If it's close to 50, recount. It's better to be at 45 than to risk going over.

5. Write the body (if needed)

Add a body when the subject line alone doesn't explain WHY the change was made. Separate from subject with a blank line.

Include a body when:

• The change is non-obvious (why this approach?)
• There's important context (related issue, design decision)
• The diff is large (summarize what changed and why)

Body rules:

• Wrap at 72 characters per line
• Explain motivation and contrast with previous behavior
• Use bullet points for multiple related changes

Skip the body when:

• The change is self-explanatory (e.g., fix(typo): correct spelling in README)
• The subject line fully captures the intent

6. Add breaking change footer (if needed)

If the commit introduces a breaking change (removes/renames a public API, changes behavior that consumers depend on), add a footer:

BREAKING CHANGE: <description of what breaks and migration path>

Rules:

• BREAKING CHANGE must be uppercase, followed by a colon and space
• Describe what breaks AND how to migrate
• Alternatively, add ! after the type/scope: feat(api)!: remove v1 endpoints
• Both notations can be used together for maximum visibility

Examples

Example 1: Simple feature addition

Diff:

diff --git a/src/utils/validators.ts b/src/utils/validators.ts
index 1a2b3c4..5d6e7f8 100644
--- a/src/utils/validators.ts
+++ b/src/utils/validators.ts
@@ -15,6 +15,18 @@ export function validateEmail(email: string): boolean {
   return EMAIL_REGEX.test(email);
 }

+export function validatePhoneNumber(phone: string): boolean {
+  // E.164 format: +[country code][number], 7-15 digits
+  const E164_REGEX = /^\+[1-9]\d{6,14}$/;
+  if (!phone) return false;
+  return E164_REGEX.test(phone);
+}

Output:

feat(validators): add phone number validation

Support E.164 format phone number validation for international
numbers. Returns false for empty strings and invalid formats.

Example 2: Bug fix with breaking change

Diff:

diff --git a/src/api/users.ts b/src/api/users.ts
--- a/src/api/users.ts
+++ b/src/api/users.ts
@@ -8,8 +8,8 @@ import { db } from '../db';
-export async function getUser(id: string): Promise<User | null> {
-  return db.users.findOne({ id });
+export async function getUser(id: string): Promise<User> {
+  const user = db.users.findOne({ id });
+  if (!user) throw new NotFoundError(`User ${id} not found`);
+  return user;
 }

Output:

fix(api): throw NotFoundError instead of returning null

getUser() previously returned null for missing users, which caused
silent failures downstream. Now throws NotFoundError with the user
ID for proper error handling.

BREAKING CHANGE: getUser() no longer returns null. Callers using
null checks must switch to try/catch with NotFoundError.

Example 3: Multi-file refactor

Diff (abbreviated):

diff --git a/src/services/payment.ts b/src/services/payment-processor.ts
similarity index 85%
rename from src/services/payment.ts
rename to src/services/payment-processor.ts

diff --git a/src/services/order.ts b/src/services/order.ts
-import { processPayment } from './payment';
+import { processPayment } from './payment-processor';

diff --git a/src/routes/checkout.ts b/src/routes/checkout.ts
-import { processPayment } from '../services/payment';
+import { processPayment } from '../services/payment-processor';

Output:

refactor(services): rename payment module to payment-processor

Rename for clarity since this module handles payment processing
specifically, not payment models or types. Updated all import
paths across order service and checkout routes.

Common mistakes

Quick reference

<type>(<scope>): <subject>     ← 50 chars max for full line
                                ← blank line
<body>                          ← 72 chars per line, explain WHY
                                ← blank line
<footer>                        ← BREAKING CHANGE: description

Types: feat | fix | refactor | docs | test | perf | style | build | ci | chore

Imperative mood test: "If applied, this commit will <subject>"

Scope: Single lowercase word for the affected module. Omit if too broad.

Key principles

1. The diff is the source of truth — Always read git diff --cached before writing. Never assume what changed based on conversation context alone.
2. Imperative mood, always — "add", "fix", "remove", "update" — never past tense or third person. This matches git's own conventions (Merge branch, Revert commit).
3. One logical change per commit — If the diff contains unrelated changes, suggest the user split them into separate commits before writing the message.
4. Subject line ≤ 50 characters — This is a hard limit, not a guideline. It ensures readability across all git tooling. Move details to the body.
5. Breaking changes must be explicit — Any change to a public API (signature, return type, removed export) requires a BREAKING CHANGE: footer. This drives semantic versioning — missing it causes silent breaking releases.