- Home
- Skills
- Code Quality & Review
- Upgrade Impact Report: What Breaks If You Update
More screenshots
Works with the AI Agents you already use
Upgrade Impact Report: What Breaks If You Update
Each step lists the packages with from and to versions, where they are used, what could break, what to test and how to roll back.
$6
Upgrade Impact Report: What Breaks If You Update
Example session with this skill installed
We are behind on our npm and Python packages. What breaks if we update? Plan it from our lockfiles, with a way back for each step.
- Read your context and instructions
- Compiled the upgrade impact report
Step 2 (careful): express 4.19.2 -> 5.1.0
Packages
express4.19.2 -> 5.1.0 (major, runtime)@types/express4.17.21 -> 5.0.0 (major, dev): the type definitions ofexpress, moved in the same step so the types match the code.
npm install express@5.1.0
npm install -D @types/express@5.0.0
Where used
src/app.ts:1: creates the app, parses JSON bodies and mounts every router.src/routes/health.ts:1,src/routes/labels.ts:1,src/routes/parcels.ts:1:Router.src/middleware/auth.ts:1,src/middleware/errors.ts:1: types only (import type).
What could break
- Question: the Express 5 release notes and migration guide were not read in this session. Who reads them and checks them against the lines in this step before merging?
What to test
npm run typecheck(package.json:14): the type definitions move in this step.npm run build(package.json:11)npm test(package.json:13):test/parcels.test.tsandtest/labels.test.tscall the API throughcreateApp.- By hand, on a staging copy:
/healthanswers, and an unknown parcel route such as/parcels/a/bstill gets the JSON 404 fromsrc/app.ts:17.
Rollback
git checkout -- package.json package-lock.json
npm ci
- Confirm:
npm ls express @types/expressshows 4.19.2 and 4.17.21, andnpm testpasses.
$ python3 scripts/plan_check.py upgrade/upgrade-plan.md --facts upgrade/facts.json --project .
summary: 0 FIX, 0 ADVICE in upgrade-plan.md (7 steps: 1 batch, 6 careful; 14 packages; 22 open questions). It checks evidence and structure against the lockfiles, manifests and source; it does not tell whether the upgrade is safe: run each step's tests on a branch.
Connects securely to your tools. The creator never sees your data.
About this skill
English
For developers who need to update dependencies without breaking the build.
Point your agent at your project. You get back an upgrade plan ordered by risk, built from your own lockfiles and source: where each package is used, what to test, and how to go back, every point with its file and line.
What you get each run
upgrade-plan.md: a summary, the risk rule, then the steps: one batch step for low-risk patch and minor updates, then careful steps one package at a time. Each step lists the packages with from and to versions, where they are used, what could break, what to test and how to roll back.upgrade-plan.html: the same as one printable page with the order of work at the top.facts.json: what the scan found, every fact with its file and line.
Read from your files
A bundled Python script (standard library, read-only, no network calls) reads package.json with package-lock.json (v2 and v3), yarn.lock v1 or pnpm-lock.yaml, and requirements*.txt or pyproject.toml with poetry.lock or uv.lock. It records each package's locked version and range with their lines, direct or transitive, runtime or dev, and every place the code imports it. Other lockfiles it finds are listed as not read.
Nothing from memory
Target versions come from the list your package manager writes with its own outdated command, or from you, and the plan names the list. A "what could break" line cites release notes that someone actually read, with the link; otherwise it is a question for the team.
Every step can be undone
Each step has the project's real test commands, with the line of the script they come from, and a rollback: restore the manifest and lockfile from git, run the clean install, and how to confirm it worked.
Checked, not just written
A checker confirms that every package and version matches the lockfile, every cited line holds the import, every step has a test and a rollback, and no destructive command appears. The agent fixes and checks again until no FIX is left.
Honest about its limits
The plan shows where the risk is; only running each step's tests on a branch shows whether an update is safe. Import sites are found by matching the source text, so imports built at runtime may be missed. Nothing is installed or run to make the plan.
Arabic
The plan can be written in Arabic with a right-to-left page; package names, versions and commands stay as they are.
In the download: SKILL.md, the scripts, references (the plan format, the inputs each parser reads and the outdated commands per package manager), and a complete example so you can see the output before you buy: a Node API with a Python service, 14 packages in 7 steps, in English and Arabic.
بالعربي
للمبرمجين الذين يحتاجون تحديث المكتبات بدون أن ينكسر المشروع.
وجّه وكيلك إلى مشروعك. تستلم خطة تحديث مرتبة بالمخاطر، مبنية من ملفات القفل والكود عندك: أين تُستعمل كل مكتبة، وماذا تختبر، وكيف ترجع، وكل نقطة مع ملفها وسطرها.
ماذا تستلم في كل مرة
- ملف
upgrade-plan.md: ملخص، وقاعدة ترتيب المخاطر، ثم الخطوات: خطوة واحدة مجمّعة لتحديثات patch وminor قليلة الخطر، ثم خطوات حذرة لكل مكتبة وحدها. كل خطوة فيها المكتبات مع الإصدار الحالي والجديد، وأين تُستعمل، وما قد ينكسر، وماذا تختبر، وكيف ترجع. - صفحة
upgrade-plan.html: الشيء نفسه كصفحة واحدة قابلة للطباعة، وترتيب العمل في أعلاها. - ملف
facts.json: ما وجده الفحص، وكل معلومة مع ملفها وسطرها.
من ملفاتك أنت
سكربت بايثون مرفق (يقرأ فقط، بلا اتصال بالشبكة) يقرأ package.json مع package-lock.json أو yarn.lock أو pnpm-lock.yaml، وrequirements*.txt أو pyproject.toml مع poetry.lock أو uv.lock. يسجّل الإصدار المقفل لكل مكتبة ونطاقها مع أسطرهما، ومباشرة أم غير مباشرة، وللتشغيل أم للتطوير، وكل مكان يستوردها فيه الكود.
لا شيء من الذاكرة
الإصدارات الجديدة تأتي من القائمة التي يكتبها مدير الحزم بأمر outdated، أو منك، والخطة تذكر مصدرها. وسطر "ما قد ينكسر" يذكر ملاحظات إصدار قرأها أحد فعلًا مع رابطها، وإلا صار سؤالًا للفريق.
كل خطوة يمكن الرجوع عنها
لكل خطوة أوامر الاختبار الحقيقية في المشروع مع سطرها، وطريقة رجوع: استرجاع ملف التعريف وملف القفل من git، ثم التثبيت النظيف، وكيف تتأكد أن الرجوع نجح.
مدقَّقة، لا مكتوبة فقط
أداة فحص تتأكد أن كل مكتبة وإصدار يطابق ملف القفل، وكل سطر مذكور فيه الاستيراد فعلًا، ولكل خطوة اختبار وطريقة رجوع، ولا يوجد أمر خطِر. يصلح الوكيل ويعيد الفحص حتى لا يبقى أي FIX.
صادقة في حدودها
الخطة تبيّن أين الخطر؛ وتشغيل اختبارات كل خطوة على فرع منفصل هو ما يبيّن هل التحديث آمن. والاستيرادات تُكتشف بمطابقة نص الكود، فالاستيراد الذي يُبنى وقت التشغيل قد يفوت.
العربية
تُكتب الخطة بالعربي في صفحة من اليمين إلى اليسار، وأسماء المكتبات والإصدارات والأوامر تبقى كما هي.
داخل الملف: ملف SKILL.md، والسكربتات، والمراجع، ومثال كامل لترى النتيجة قبل الشراء: واجهة Node مع خدمة بايثون، 14 مكتبة في 7 خطوات، بالإنجليزي والعربي.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Code Quality & Review.
Reads package.json with package-lock.json (v2, v3), yarn.lock v1 or pnpm-lock.yaml, and requirements*.txt or pyproject.toml with poetry.lock or uv.lock; other lockfiles (Go, Rust, Ruby, PHP and others) are listed as found, not read. Target versions come from your package manager's outdated list or from you, never from memory. Breaking changes are cited only from release notes someone read; otherwise they are questions. Import sites are found by matching the source text, so imports built at runtime may be missed. The plan shows where the risk is; only running each step's tests shows whether an update is safe. upgrade-plan.html loads its fonts from Google Fonts when online and falls back to system fonts offline. الخطة تبيّن أين الخطر، والاختبارات هي التي تبيّن هل التحديث آمن.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 3 days ago
- Software
Python
Gittsc
Node.js
Permissions
Reads the project's manifests, lockfiles and source files, and writes facts.json, upgrade-plan.md and upgrade-plan.html in an upgrade folder. The agent runs one command itself, after telling you: the package manager's own outdated listing, saved to a file. The bundled scripts (deps_scan.py, plan_check.py, make_page.py) run nothing, install nothing and make no network calls. يقرأ ملفات المشروع ويكتب ملفات الخطة فقط، ولا يثبّت شيئًا.