Eval npm

NPM Package Evaluator

Evaluate npm packages to find the best option for a given use case. Compare maintenance health, community adoption, and identify promising newcomers while avoiding abandoned projects.

Instructions

Step 0: Security Advisory Pre-Check

Before evaluating any packages, run the npm-security-advisory skill procedure against the target packages. The skill produces three classes of findings that factor into evaluation:

1. Advisory hits from Socket.dev/OSV/npm audit (known CVEs, known malware).
2. Step 4 metadata anomalies — supply-chain worm signals: FRESH, SCRIPT_ADDED, CADENCE_BURST, PROVENANCE_REGRESSION, NEW_PUBLISHER+FRESH, IOC_MATCH, YANKED, etc. These catch attacks before threat feeds do.
3. Positive signals like PROVENANCE_OK — surface these in the report as confidence-raising evidence.

• Mode 1 (audit): Check all packages from package.json. MALWARE/CRITICAL findings (advisory hit OR IOC_MATCH OR two combined HIGH Step 4 signals) automatically score 0 on the "Quality Signals: known security vulnerabilities" dimension and get flagged with a red flag. Individual HIGH Step 4 signals knock the dimension down by 3 points each. PROVENANCE_OK adds 1 point. Do not skip evaluation — the user needs the full picture to choose a replacement.
• Mode 2 (compare): Check all candidate packages. Advisory findings and Step 4 signals are surfaced in the comparison table — add a "Supply-chain signals" row alongside the existing "Security" row. A package with active MALWARE findings (advisory or IOC_MATCH) gets an automatic "DO NOT USE" recommendation. A package with PROVENANCE_OK gets a small ranking boost when alternatives are otherwise tied.

This command supports two modes:

Mode 1: Package.json Audit

If the user says "audit my package.json", "check my dependencies", "scan package.json", or provides a package.json file:

1. Read the package.json from the current directory or the provided file

2. Extract all dependencies from both dependencies and devDependencies

3. Triage scan each package - do a quick health check (not full evaluation):

   • Last publish date
   • Weekly downloads trend
   • Known vulnerabilities
   • Maintenance status (active/maintenance-mode/abandoned)

4. Categorize results:

For any packages flagged as problematic, use npmtrends to find alternatives:

• Search [package-name] alternatives site:npmtrends.com or
• Fetch https://npmtrends.com/[flagged-pkg]-vs-[known-alternative] to compare

## Package.json Audit Results

### 🚨 Immediate Attention (Abandoned/Vulnerable)
Packages that are security risks or completely unmaintained:
| Package | Issue | Last Publish | Action |
|---------|-------|--------------|--------|
| left-pad | Abandoned | 3 years ago | Replace with native `.padStart()` |
| vulnerable-pkg | 2 CVEs | 6 months ago | Update or replace |

### ⚠️ Review Recommended (Aging/Declining)
Packages showing warning signs:
| Package | Concern | Better Alternative | Migration Effort |
|---------|---------|-------------------|------------------|
| moment | Maintenance mode | date-fns | Medium |
| request | Deprecated | got, axios, fetch | Medium |

### ✅ Healthy (No Action Needed)
[List packages that are actively maintained with no concerns]

### 📊 Summary
- Total packages: X
- Healthy: X (X%)
- Needs review: X (X%)
- Critical: X (X%)

### 🎯 Recommended Actions (Priority Order)
1. **URGENT**: [package] - [specific action]
2. **Soon**: [package] - [specific action]
3. **When convenient**: [package] - [specific action]

1. Offer deep-dive: "Want me to do a full evaluation on any of these packages?"

Efficiency notes for large projects:

• For projects with 30+ dependencies, batch searches (e.g., "[pkg1] [pkg2] [pkg3] npm maintenance status")
• Prioritize scanning: check obviously risky packages first (old, low downloads, unfamiliar names)
• Skip well-known healthy packages (react, typescript, eslint, etc.) unless specifically asked
• Focus effort on packages in dependencies over devDependencies (runtime risk > build risk)

Mode 2: Package Evaluation

When the user provides package names or a category/problem to solve, perform a comprehensive evaluation:

Step 1: Gather Package Candidates

If given a category (e.g., "date formatting", "state management"), search for the top packages in that space.

If given specific package names, evaluate those BUT ALSO:

• Search for "best alternatives to [package]" or "[package] vs"
• Identify the top 2-3 competitors in the same problem space
• Include any newer packages gaining traction
• Always expand a single package query into a comparison - the user needs context to make a good decision

For example:

• User asks about moment → also evaluate date-fns, dayjs, luxon
• User asks about express → also evaluate fastify, hono, elysia
• User asks about lodash → also evaluate radash, remeda, or native alternatives

Step 2: Research Each Package

Primary data source: npmtrends.com

For package comparisons, fetch https://npmtrends.com/[pkg1]-vs-[pkg2]-vs-[pkg3] to get:

• Download trend graphs (visual trajectory over 1/6/12/24 months)
• Side-by-side stats table (stars, issues, updated, created, size, dependencies)
• Direct comparison of momentum

Example URLs to fetch:

• https://npmtrends.com/date-fns-vs-dayjs-vs-moment-vs-luxon
• https://npmtrends.com/express-vs-fastify-vs-hono-vs-koa
• https://npmtrends.com/prisma-vs-drizzle-orm-vs-kysely

Always fetch npmtrends first - it gives you most of the data in one request. Then supplement with additional searches for:

For each package, gather:

Maintenance Health (40% weight)

• Last publish date on npm (red flag if >1 year)
• Commit frequency in the last 6 months
• Open vs closed issues ratio
• PR merge velocity
• Number of active maintainers (bus factor risk if 1)
• Whether it's org-backed or solo maintained

Community Adoption (30% weight)

• Weekly npm downloads (and trend direction)
• GitHub stars
• Number of dependent packages (npm dependents)
• Stack Overflow question volume
• Discord/community presence

Quality Signals (20% weight)

• TypeScript support: native | @types | none
• Bundle size (check bundlephobia.com for frontend packages)
• Known security vulnerabilities
• Dependency count (prefer minimal)
• Documentation quality and examples
• Test coverage if visible

Trajectory & Momentum (10% weight) - npmtrends is the primary source for this

• Download trend graph: growing | stable | declining (look at 6-12 month view)
• Crossover points: is a newer package about to overtake an established one?
• Star velocity (recent stars vs total)
• Recent major releases or active roadmap
• Newer alternatives gaining traction

Step 3: Red Flags to Identify

Explicitly call out packages with:

• ❌ No commits in 12+ months
• ❌ No npm publish in 12+ months
• ❌ Unaddressed security vulnerabilities
• ❌ Single maintainer who appears inactive
• ❌ Declining download trends
• ❌ Archived repository
• ❌ No TypeScript support (for TS projects)
• ❌ Massive bundle size relative to alternatives
• ❌ Excessive dependency tree

Step 4: Identify Promising Newcomers

Use npmtrends to spot rising packages:

• Look for steep upward curves on newer packages
• Watch for "crossover moments" where a new package overtakes an established one
• Check if a package went from 0 to 100k+ downloads in under 2 years

Look for newer packages that:

• ✨ Were created in the last 2 years
• ✨ Show strong growth trajectory
• ✨ Are backed by known developers/companies
• ✨ Address pain points of established alternatives
• ✨ Have modern architecture (ESM, tree-shakeable, etc.)

Step 5: Deliver Structured Output

Present findings in this format:

## Package Evaluation: [Category/Use Case]

### Quick Verdict
[1-2 sentence recommendation]

**📈 View trends**: [npmtrends.com/pkg1-vs-pkg2-vs-pkg3](link)

### Score Summary

Rate each package 1-10 and calculate weighted score:

| Package | Maintenance (40%) | Adoption (30%) | Quality (20%) | Trajectory (10%) | **Total** |
|---------|-------------------|----------------|---------------|------------------|-----------|
| queried | 6                 | 8              | 7             | 4                | **6.4**   |
| alt-1   | 9                 | 7              | 9             | 8                | **8.2** ⭐ |
| alt-2   | 8                 | 6              | 8             | 9                | **7.5**   |

**If an alternative scores higher than the package the user asked about, lead with that finding.**

### Comparison Matrix

| Package | Downloads/wk | Stars | Last Publish | TS Support | Bundle | Verdict |
|---------|-------------|-------|--------------|------------|--------|---------|
| pkg-a   | 2.1M        | 45k   | 2 weeks ago  | Native     | 12kb   | ✅ Best |
| pkg-b   | 890k        | 23k   | 8 months ago | @types     | 45kb   | ⚠️ Aging |
| pkg-c   | 150k        | 5k    | 1 week ago   | Native     | 8kb    | 🌟 Rising |

### Detailed Analysis

#### [Package Name] - [Verdict Emoji] [One-word status]
- **Maintenance**: [assessment]
- **Adoption**: [assessment]
- **Quality**: [assessment]
- **Trajectory**: [assessment]
- **Best for**: [use case fit]
- **Watch out for**: [concerns]

### Recommendation

**🏆 Best Overall**: [package] because [reason]

**⚠️ Better Alternatives Exist**: If evaluating a package that scores lower than alternatives, prominently call this out:
> You asked about **[package]**, but **[better-package]** scores higher on [metrics]. Consider switching because [reasons]. Migration effort: [low/medium/high].

**For most projects**: [package] because [reason]

**If you need [specific thing]**: Consider [alternative] because [reason]

**Avoid**: [package] because [reason]

**Keep an eye on**: [newer package] which [promise]

Step 6: Answer Follow-up Questions

Be prepared to:

• Deep dive on any specific package
• Compare specific features
• Check compatibility with user's stack
• Find migration guides between packages
• Identify packages for related needs

Usage Examples

Audit mode:

• "audit my package.json"
• "check my dependencies for problems"
• "scan dependencies" (will look for package.json in current directory)
• "audit this package.json: [paste contents]"
• "which of my packages are abandoned?"

Evaluation mode:

• "Evaluate date libraries: date-fns vs dayjs vs luxon vs moment"
• "What's the best React form library in 2024?"
• "Compare ORMs for Node.js with Postgres"
• "Find me a lightweight carousel package"
• "Is lodash still worth using?"

$ARGUMENTS