feat: add automated PR review script

- Adds `scripts/agent/review-pr.mjs` to automate PR checks (lint, format, test).
- Implements micro-fix logic to auto-commit formatting changes.
- Adds guardrails for sensitive files (security/protocol review).
- Generates a Markdown report with test failure analysis.
This commit is contained in:
google-labs-jules[bot] committed 2026-01-20 23:29:02 +00:00
1 parent f2fd3c42d4
commit c2b0f713e5
2 files changed
+49 -97

No files matched your search

-46
View File
@@ -1,46 +0,0 @@
# Automated PR Review
**Context:** jules-17818997485755159011-27e733a1
**Date:** 2026-01-19T23:31:08.588Z
## ✅ Formatting
No formatting issues found.
## ⚠️ Linter Warnings/Errors
`npm run lint` reported issues:
<details><summary>Show Lint Output</summary>
```
> [email protected] lint
> npm run lint:css && npm run lint:hex && npm run lint:inline-styles && npm run lint:tokens && npm run lint:tailwind-brackets && npm run lint:tailwind-colors
> [email protected] lint:css
> stylelint css/tailwind.source.css css/tokens.css
> [email protected] lint:hex
> node scripts/check-hex.js
> [email protected] lint:inline-styles
> node scripts/check-inline-styles.mjs
No inline style usage found.
> [email protected] lint:tokens
> node scripts/check-design-tokens.mjs --check=tokens
Design token lint failed: raw measurements detected outside tokens.
css/tailwind.source.css:5598 → 2rem
min-width: 2rem;
css/tailwind.source.css:5613 → 8rem
min-width: 8rem;
Move these measurements into css/tokens.css or use the metrics helper.
```
</details>
## ❌ Test Timeout
`npm run test:unit` timed out after 5 minutes.
+49 -51
View File
@@ -1,4 +1,4 @@
import { execSync, spawnSync } from 'child_process';
import { spawnSync } from 'child_process';
import fs from 'fs';
import path from 'path';
@@ -19,9 +19,7 @@ function runCommand(command, args = [], options = {}) {
return spawnSync(command, args, {
encoding: 'utf-8',
stdio: ['ignore', 'pipe', 'pipe'],
shell: false, // Turn off shell for security unless needed, but spawnSync defaults to false.
// However, if I used shell: true before, args might need care.
// If shell is false, args are safe.
shell: false,
...options
});
}
@@ -38,17 +36,21 @@ function checkGhAvailability() {
function getChangedFiles(baseBranch = 'origin/main') {
try {
const res = spawnSync('git', ['diff', '--name-only', `${baseBranch}...HEAD`], { encoding: 'utf-8' });
if (res.status !== 0) throw new Error(res.stderr);
if (res.status !== 0) {
// Fallback: use local changes
const local = spawnSync('git', ['diff', '--name-only', 'HEAD'], { encoding: 'utf-8' });
return local.stdout ? local.stdout.trim().split('\n').filter(Boolean) : [];
}
return res.stdout.trim().split('\n').filter(Boolean);
} catch (error) {
log(`Warning: Could not determine changed files against ${baseBranch}. Using local changes if any.`);
log(`Warning: Could not determine changed files. Using local changes if any.`);
return [];
}
}
async function reviewContext(contextName, options) {
const report = [];
report.push(`# Automated PR Review`);
report.push(`# Automated PR Review & Suggested Improvements`);
report.push(`**Context:** ${contextName}`);
report.push(`**Date:** ${new Date().toISOString()}`);
report.push('');
@@ -59,9 +61,6 @@ async function reviewContext(contextName, options) {
// 1. Install dependencies
if (!options.noInstall) {
log('Installing dependencies (npm ci)...');
// On Windows, npm is a cmd. Using shell: true might be needed for 'npm' command,
// or use 'npm.cmd'. Linux is fine.
// Safe to use shell: true for static commands.
const ciResult = spawnSync('npm', ['ci'], { encoding: 'utf-8', shell: true });
if (ciResult.status !== 0) {
report.push(`## 🚨 Dependency Installation Failed`);
@@ -86,19 +85,20 @@ async function reviewContext(contextName, options) {
if (formattedFiles.length > 0) {
report.push(`## 🎨 Formatting Suggestions`);
report.push(`The following files need formatting. Run \`npm run format\` to fix.`);
report.push(`The following files need formatting:`);
report.push('<ul>');
formattedFiles.forEach(f => report.push(`<li>${f}</li>`));
report.push('</ul>');
report.push('');
// Fast-path suggestion
report.push('**Suggestion:** Apply `npm run format` and re-run tests.');
report.push('**Suggestion:** apply `npm run format` and re-run tests');
if (options.applyFixes) {
try {
log('Applying formatting fixes...');
runCommand('git', ['add', '.']);
// Explicit wording: fix(ai): <short> (agent)
runCommand('git', ['commit', '-m', 'fix(ai): formatting (agent)']);
log('Pushing changes...');
@@ -111,44 +111,32 @@ async function reviewContext(contextName, options) {
report.push('**Action:** Applied formatting fixes locally, but failed to push.');
report.push(`**Error:** ${pushResult.stderr}`);
// Attempt fallback: Create new branch?
// If we are in a detached HEAD or locked branch, pushing might fail.
// "open a follow-up PR if you could not directly update the PR branch"
// If we have GH CLI...
if (checkGhAvailability()) {
const newBranch = `fix/ai-formatting-${Date.now()}`;
log(`Attempting fallback: creating new branch ${newBranch}...`);
runCommand('git', ['checkout', '-b', newBranch]);
const pushNew = runCommand('git', ['push', '-u', 'origin', newBranch]);
if (pushNew.status === 0) {
// Create PR
log('Creating fallback PR...');
// We need to target the original branch? Or main?
// Let's target the current context if possible.
// But we don't know the original target easily without GH CLI.
// We'll just create a PR against default.
const prRes = spawnSync('gh', ['pr', 'create', '--title', 'fix(ai): formatting (agent)', '--body', 'Automated formatting fixes.', '--fill'], { encoding: 'utf-8' });
if (prRes.status === 0) {
report.push(`**Action:** Created follow-up PR: ${prRes.stdout.trim()}`);
} else {
report.push(`**Error:** Failed to create follow-up PR: ${prRes.stderr}`);
}
} else {
report.push(`**Error:** Failed to push fallback branch: ${pushNew.stderr}`);
}
// Switch back?
// reviewContext usually assumes we stay on the branch or restore.
}
}
} catch (e) {
log(`Error applying fixes: ${e.message}`);
report.push(`**Error:** Failed to apply formatting fixes: ${e.message}`);
}
} else {
// Revert changes so checking other things is clean?
// If we don't apply fixes, we should revert.
// If we're not applying fixes, we should revert the formatting changes so they don't affect subsequent steps?
// However, keeping them might be good for running tests on clean code?
// But if tests rely on formatting (unlikely), it matters.
// Usually we revert so the working tree matches the PR state for further analysis.
spawnSync('git', ['checkout', '.'], { stdio: 'ignore' });
}
} else {
report.push(`## ✅ Formatting`);
@@ -163,11 +151,21 @@ async function reviewContext(contextName, options) {
if (lintResult.status !== 0) {
report.push(`## ⚠️ Linter Warnings/Errors`);
report.push('`npm run lint` reported issues:');
// Extract lines with "Error" or "Warning"
const lines = (lintResult.stdout + '\n' + lintResult.stderr).split('\n');
const issues = lines.filter(l => l.match(/error|warning/i)).slice(0, 20); // Limit to 20
report.push('<details><summary>Show Lint Output</summary>');
report.push('');
report.push('```');
report.push(lintResult.stdout);
report.push(lintResult.stderr);
if (issues.length > 0) {
report.push(issues.join('\n'));
if (lines.length > issues.length) report.push('...');
} else {
report.push(lintResult.stdout.slice(-1000));
report.push(lintResult.stderr.slice(-1000));
}
report.push('```');
report.push('</details>');
@@ -189,12 +187,27 @@ async function reviewContext(contextName, options) {
report.push('`npm run test:unit` timed out after 5 minutes.');
} else if (testResult.status !== 0) {
report.push(`## ❌ Test Failures`);
report.push('`npm run test:unit` failed. Inspect the stack traces below.');
report.push('`npm run test:unit` failed.');
// Analyze stdout for failing test file
// Expecting "→ Running tests/..." lines.
const output = testResult.stdout || '';
const runningLines = output.match(/→ Running (.*)/g);
let lastTest = 'Unknown';
if (runningLines && runningLines.length > 0) {
lastTest = runningLines[runningLines.length - 1].replace('→ Running ', '').trim();
}
report.push(`**Possible culprit:** \`${lastTest}\``);
report.push('Suggested area to inspect: check the stack trace below for the failing assertion.');
report.push('<details><summary>Show Test Output</summary>');
report.push('');
report.push('```');
report.push(testResult.stdout ? testResult.stdout.slice(-5000) : 'No stdout');
report.push(testResult.stderr ? testResult.stderr.slice(-5000) : 'No stderr');
// Capture context around the failure?
// Usually the end of stdout has the failure.
report.push(output.slice(-3000));
report.push((testResult.stderr || '').slice(-3000));
report.push('```');
report.push('</details>');
} else {
@@ -211,19 +224,13 @@ async function reviewContext(contextName, options) {
if (sensitiveChanges.length > 0) {
report.push(`## 🔒 Security Review Required`);
report.push(`This PR modifies sensitive files. **requires-security-review**, **requires-protocol-review**`);
report.push(`This PR modifies sensitive files.`);
report.push(`**Flags:** \`requires-security-review\`, \`requires-protocol-review\``);
report.push('<ul>');
sensitiveChanges.forEach(f => report.push(`<li>${f}</li>`));
report.push('</ul>');
}
// Cleanup formatting if not applied
if (!options.applyFixes) {
try {
spawnSync('git', ['checkout', '.'], { stdio: 'ignore' });
} catch (e) {}
}
return report.join('\n');
} catch (error) {
@@ -242,7 +249,6 @@ async function main() {
};
const branchArg = args.find(a => !a.startsWith('--') && args.indexOf(a) !== args.indexOf('--pr') + 1);
const ghAvailable = checkGhAvailability();
if (options.all) {
@@ -250,7 +256,6 @@ async function main() {
console.error('Error: `gh` CLI not found. Cannot list open PRs.');
process.exit(1);
}
// Fetch PRs
log('Fetching open PRs...');
try {
const prsJson = spawnSync('gh', ['pr', 'list', '--json', 'number,headRefName,url'], { encoding: 'utf-8' });
@@ -263,14 +268,11 @@ async function main() {
try {
spawnSync('gh', ['pr', 'checkout', pr.number]);
const report = await reviewContext(`PR #${pr.number}`, options);
// Post comment
const tempFile = `pr_review_${pr.number}.md`;
fs.writeFileSync(tempFile, report);
log(`Posting comment to PR #${pr.number}...`);
spawnSync('gh', ['pr', 'comment', pr.number, '-F', tempFile]);
fs.unlinkSync(tempFile);
} catch (e) {
log(`Failed to process PR #${pr.number}: ${e.message}`);
}
@@ -288,19 +290,15 @@ async function main() {
try {
const res = spawnSync('gh', ['pr', 'checkout', options.pr], { encoding: 'utf-8' });
if (res.status !== 0) throw new Error(res.stderr);
const report = await reviewContext(`PR #${options.pr}`, options);
console.log(report);
fs.writeFileSync(LOG_FILE, report);
log(`Report saved to ${LOG_FILE}`);
} catch (e) {
console.error(e);
process.exit(1);
}
} else {
// Current branch or specified branch
let originalBranch;
if (branchArg) {
try {