content-grade 1.0.0 → 1.0.1

package/README.md

@@ -96,8 +96,17 @@ npx content-grade headline "Why Most Startups Fail at Month 18"
 | `grade "<text>"` | Alias for `headline` |
 | `init` | First-run setup: verify Claude CLI, run smoke test |
 | `start` | Launch the full web dashboard (6 tools) |
+| `telemetry [on\|off]` | View or toggle anonymous usage tracking |
 | `help` | Full usage and examples |
 
+**Global flags:**
+
+| Flag | Description |
+|------|-------------|
+| `--no-telemetry` | Skip usage tracking for this one invocation |
+| `--version`, `-v`, `-V` | Show installed version |
+| `--help`, `-h` | Show help |
+
 ---
 
 ## The 6 web dashboard tools
@@ -207,6 +216,26 @@ Use this when something isn't working — it tells you exactly what to fix.
 
 ---
 
+### `telemetry`
+
+```bash
+content-grade telemetry          # show current status
+content-grade telemetry off      # opt out — persisted permanently
+content-grade telemetry on       # opt back in
+```
+
+Anonymous usage tracking (command name + score range, no file contents, no PII). Enabled by default; `telemetry off` writes a local flag file and is honoured by all future invocations.
+
+**One-time opt-out per invocation:**
+
+```bash
+content-grade analyze ./post.md --no-telemetry
+```
+
+Pass `--no-telemetry` to any command to skip tracking for that run only without changing the saved preference.
+
+---
+
 ### `start` / `serve`
 
 ```bash
@@ -312,6 +341,353 @@ npm run test:coverage # coverage report
 
 ---
 
+## Output format
+
+When `content-grade analyze` runs, Claude returns structured JSON that is rendered to the terminal. The raw schema:
+
+```json
+{
+  "content_type": "blog | email | ad | social | landing",
+  "total_score": 72,
+  "grade": "B+",
+  "headline": {
+    "text": "Why Most Startups Fail",
+    "score": 61,
+    "feedback": "Too vague — the reader doesn't know if this applies to them."
+  },
+  "dimensions": {
+    "clarity":        { "score": 80, "note": "Well-organized with clear section breaks." },
+    "engagement":     { "score": 60, "note": "Opens with a statistic but drops momentum in paragraph 3." },
+    "structure":      { "score": 72, "note": "Good use of subheadings but the conclusion is thin." },
+    "value_delivery": { "score": 65, "note": "Core advice is solid but needs a concrete takeaway." }
+  },
+  "one_line_verdict": "Strong structure undermined by a weak headline — fix that first.",
+  "strengths": [
+    "Well-organized with clear section breaks",
+    "Specific statistics in the opening hook"
+  ],
+  "improvements": [
+    {
+      "issue": "Headline is too vague",
+      "priority": "high",
+      "fix": "Why 90% of SaaS Startups Fail at Month 18 (and How to Be the 10%)"
+    },
+    {
+      "issue": "Paragraph 2 makes claims without evidence",
+      "priority": "medium",
+      "fix": "Cite a specific study — 'CB Insights found...' beats 'many startups...'"
+    }
+  ],
+  "headline_rewrites": [
+    "Why 90% of SaaS Startups Fail at Month 18 (and How to Be the 10%)",
+    "The Month 18 Startup Trap: What Kills Growth-Stage Companies"
+  ]
+}
+```
+
+For `content-grade headline`, the schema:
+
+```json
+{
+  "score": 74,
+  "grade": "B",
+  "verdict": "Strong promise but lacks the proof element that separates a 74 from an 88.",
+  "scores": {
+    "rule_of_one":        { "score": 22, "max": 30, "note": "Single idea present but promise is vague." },
+    "value_equation":     { "score": 22, "max": 30, "note": "Dream outcome clear; likelihood undermined by '10x' without proof." },
+    "readability":        { "score": 16, "max": 20, "note": "Short, no jargon — reads well." },
+    "proof_promise_plan": { "score": 14, "max": 20, "note": "Timeframe adds proof; missing the 'how'." }
+  },
+  "rewrites": [
+    "How I Doubled Conversions in 30 Days by Fixing One Headline",
+    "The 3 Headline Fixes That Grew My Conversions 10x (Backed by 847 A/B Tests)"
+  ]
+}
+```
+
+---
+
+## Scoring methodology
+
+ContentGrade uses Claude as a scoring engine with **calibrated anchor examples** that force use of the full 0–100 range. This prevents the common AI problem of all scores bunching in the 60–80 band.
+
+**Score thresholds (for `analyze`):**
+
+| Score | What it means |
+|-------|--------------|
+| 90–100 | Exceptional — data authority, curiosity gap, ultra-specific, publication-ready |
+| 80–89 | Strong — clear value equation, good proof, specific promise |
+| 65–79 | Good — readable, structured, some specificity gaps |
+| 50–64 | Adequate — readable but vague, generic structure |
+| 35–49 | Poor — clichés, no hook, commodity topic |
+| 0–34 | Very weak — no structure, zero reader benefit, or content too short |
+
+**Dimensions:**
+
+- **Clarity** — Organization, section breaks, readability. Does the reader always know where they are?
+- **Engagement** — Hook strength, momentum, reader retention. Would someone read to the end?
+- **Structure** — Heading hierarchy, flow, conclusion strength. Is there a logical arc?
+- **Value Delivery** — Concrete advice, takeaways, actionable insights. Does the reader leave with something useful?
+
+**Content-type calibration:** The scoring prompt knows it's reading a blog post vs. an email vs. ad copy. A 200-word email doesn't get penalized for "thin structure" the way a blog post would.
+
+**Headline framework calibration anchors:**
+
+| Score range | What it takes |
+|-------------|--------------|
+| 95+ | Data authority + curiosity gap + ultra-specificity + clear reader identity |
+| 88–92 | Strong proof + promise + plan, low jargon, 6–10 words |
+| 75–84 | Clear value equation, objection handling, some specificity |
+| 65–74 | Clear differentiator or positioning, but vague on proof |
+| 50–64 | Decent structure but generic or lacking urgency |
+| 35–49 | Standard cliché phrasing, no hook element |
+| 0–34 | Commodity topic, passive voice, zero benefit stated |
+
+---
+
+## Programmatic use (REST API)
+
+Start the server with `content-grade start`, then call any tool via HTTP:
+
+### Grade a headline
+
+```bash
+curl -s -X POST http://localhost:4000/api/demos/headline-grader \
+  -H 'Content-Type: application/json' \
+  -d '{"headline": "Why Most Startups Fail at Month 18"}' | jq .
+```
+
+Response:
+```json
+{
+  "total_score": 74,
+  "grade": "B",
+  "diagnosis": "Strong promise but lacks the proof element that separates a 74 from an 88.",
+  "framework_scores": {
+    "rule_of_one":        { "score": 22, "max": 30, "note": "..." },
+    "value_equation":     { "score": 22, "max": 30, "note": "..." },
+    "readability":        { "score": 16, "max": 20, "note": "..." },
+    "proof_promise_plan": { "score": 14, "max": 20, "note": "..." }
+  },
+  "rewrites": ["...", "..."],
+  "usage": { "remaining": 2, "limit": 3, "isPro": false }
+}
+```
+
+### Compare two headlines (Pro)
+
+```bash
+curl -s -X POST http://localhost:4000/api/demos/headline-grader/compare \
+  -H 'Content-Type: application/json' \
+  -d '{
+    "headlineA": "Why Most Startups Fail",
+    "headlineB": "Why 90% of SaaS Startups Fail at Month 18",
+    "email": "you@example.com"
+  }' | jq .
+```
+
+### Audit a landing page URL (Pro)
+
+```bash
+curl -s -X POST http://localhost:4000/api/demos/page-roast \
+  -H 'Content-Type: application/json' \
+  -d '{"url": "https://example.com", "email": "you@example.com"}' | jq .
+```
+
+### Score ad copy
+
+```bash
+curl -s -X POST http://localhost:4000/api/demos/ad-scorer \
+  -H 'Content-Type: application/json' \
+  -d '{
+    "adCopy": "Stop wasting money on ads that don'\''t convert.",
+    "platform": "facebook",
+    "email": "you@example.com"
+  }' | jq .
+```
+
+### Grade a Twitter thread
+
+```bash
+curl -s -X POST http://localhost:4000/api/demos/thread-grader \
+  -H 'Content-Type: application/json' \
+  -d '{
+    "threadText": "1/ Most people think productivity is about working more hours...\n\n2/ They'\''re wrong.",
+    "email": "you@example.com"
+  }' | jq .
+```
+
+### Generate an email
+
+```bash
+curl -s -X POST http://localhost:4000/api/demos/email-forge \
+  -H 'Content-Type: application/json' \
+  -d '{
+    "product": "A CLI tool that grades content quality",
+    "audience": "developers who write technical blog posts",
+    "goal": "cold_outreach",
+    "tone": "casual",
+    "email": "you@example.com"
+  }' | jq .
+```
+
+**Rate limit response** (when daily free limit is hit):
+```json
+{
+  "gated": true,
+  "isPro": false,
+  "remaining": 0,
+  "limit": 3,
+  "message": "Free daily limit reached (3/day). Get 100 analyses/day with Pro"
+}
+```
+
+See [`docs/api.md`](./docs/api.md) for the full reference — every endpoint, complete request/response shapes, and more cURL examples.
+
+---
+
+## Workflows & integrations
+
+### GitHub Actions — CI content quality gate
+
+Fail the build if content quality drops below a threshold. Useful for content-heavy repos (marketing site, docs, blog).
+
+```yaml
+# .github/workflows/content-quality.yml
+name: Content quality check
+
+on:
+  pull_request:
+    paths:
+      - 'content/**'
+      - 'blog/**'
+      - '*.md'
+
+jobs:
+  grade:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Install Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: '20'
+
+      - name: Install Claude CLI
+        run: npm install -g @anthropic-ai/claude-code
+
+      - name: Login to Claude (non-interactive)
+        run: echo "${{ secrets.ANTHROPIC_API_KEY }}" | claude login --api-key
+        # Note: Claude CLI requires authentication. Store your API key in GitHub secrets.
+
+      - name: Grade changed content
+        run: |
+          for f in $(git diff --name-only origin/main...HEAD | grep -E '\.(md|txt|mdx)$'); do
+            echo "=== Grading $f ==="
+            npx content-grade analyze "$f" --no-telemetry
+          done
+```
+
+> **Note:** Claude CLI must be authenticated. In CI, this requires an Anthropic API key. Set `ANTHROPIC_API_KEY` in your repository secrets and use it to log in.
+
+---
+
+### Pre-commit hook
+
+Grade content before it enters the repo. Add this to `.git/hooks/pre-commit` (make it executable with `chmod +x .git/hooks/pre-commit`):
+
+```bash
+#!/bin/sh
+# Content quality pre-commit hook
+# Grades staged .md/.txt files and blocks commit if any score below threshold
+
+THRESHOLD=40
+
+staged=$(git diff --cached --name-only --diff-filter=ACMR | grep -E '\.(md|txt|mdx)$')
+
+if [ -z "$staged" ]; then
+  exit 0
+fi
+
+echo "Running ContentGrade on staged files..."
+
+for file in $staged; do
+  # Only check files over 100 characters (skip tiny files)
+  chars=$(wc -c < "$file" 2>/dev/null || echo 0)
+  if [ "$chars" -lt 100 ]; then
+    continue
+  fi
+
+  echo "  Grading $file..."
+  npx content-grade analyze "$file" --no-telemetry
+done
+
+echo "✓ Content quality check complete."
+exit 0
+```
+
+To block low-scoring commits, parse the score from output and compare against `$THRESHOLD`. The current output format prints `OVERALL SCORE   72/100` — grep for the number and exit 1 if below threshold.
+
+---
+
+### Batch processing a directory
+
+Grade every content file in a directory and save results:
+
+```bash
+# Grade all markdown files and save output
+find ./content -name '*.md' -not -name 'README.md' | while read f; do
+  echo "--- $f ---"
+  npx content-grade analyze "$f" --no-telemetry
+  echo ""
+done > content-grades-$(date +%Y%m%d).txt
+
+# Or use the built-in directory scan (picks the best candidate):
+npx content-grade ./content
+```
+
+---
+
+### Comparing scores over time
+
+Track score progression across revisions:
+
+```bash
+# Score the current version of a post
+npx content-grade analyze ./blog/my-post.md --no-telemetry > scores/my-post-$(date +%Y%m%d).txt
+
+# Compare two versions side-by-side
+diff scores/my-post-20240101.txt scores/my-post-20240201.txt
+
+# Quick numeric comparison
+grep "OVERALL SCORE" scores/my-post-*.txt
+```
+
+For structured tracking, use the REST API (run `content-grade start` first):
+
+```bash
+# Score via API, extract just the number
+curl -s -X POST http://localhost:4000/api/demos/headline-grader \
+  -H 'Content-Type: application/json' \
+  -d '{"headline": "Your Post Title Here"}' \
+  | jq '.total_score'
+```
+
+Log to a CSV:
+
+```bash
+DATE=$(date +%Y-%m-%d)
+HEADLINE="Your Post Title Here"
+SCORE=$(curl -s -X POST http://localhost:4000/api/demos/headline-grader \
+  -H 'Content-Type: application/json' \
+  -d "{\"headline\": \"$HEADLINE\"}" | jq '.total_score')
+
+echo "$DATE,$HEADLINE,$SCORE" >> headline-scores.csv
+```
+
+---
+
 ## Pricing
 
 | | Free | Pro |

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "content-grade",
-  "version": "1.0.0",
+  "version": "1.0.1",
   "description": "AI-powered content analysis CLI. Score any blog post, landing page, or ad copy in under 30 seconds — runs on Claude CLI, no API key needed.",
   "type": "module",
   "bin": {
@@ -50,7 +50,12 @@
     "test:watch": "vitest",
     "test:coverage": "vitest run --coverage",
     "stats": "node scripts/npm-stats.js",
-    "stats:week": "node scripts/npm-stats.js --week"
+    "stats:week": "node scripts/npm-stats.js --week",
+    "metrics": "node scripts/metrics.js",
+    "metrics:save": "node scripts/metrics.js --save",
+    "metrics:history": "node scripts/metrics.js --history",
+    "metrics:csv": "node scripts/metrics.js --csv",
+    "github-traffic": "node scripts/github-traffic.js"
   },
   "dependencies": {
     "@fastify/cors": "^9.0.1",