Email Subject Line Tester
by @mguozhen
Email subject line optimization agent. Generates 10 subject line variants for any email, scores each by open rate predictors (urgency, personalization, curio...
clawhub install email-subject-line-testerπ About This Skill
name: email-subject-line-tester description: "Email subject line optimization agent. Generates 10 subject line variants for any email, scores each by open rate predictors (urgency, personalization, curiosity, length, emoji use), and recommends the top 3 for A/B testing. Triggers: subject line, email subject, subject line tester, subject line generator, email open rate, ab test email, subject line optimization, email copywriting, subject line ideas, newsletter subject, email marketing subject" allowed-tools: Bash metadata: openclaw: homepage: https://github.com/mguozhen/email-subject-line-tester
Email Subject Line Tester
AI-powered email subject line optimization agent β generates 10 variants, scores each on 7 open-rate predictor signals, and selects the top 3 for A/B testing.
Describe your email's topic, audience, and goal. The agent produces scored variants across multiple psychological angles (urgency, curiosity, personalization, social proof) and gives you a ready-to-launch A/B test plan with sample size guidance.
Commands
subject test # generate and score 10 subject line variants for a topic
subject generate # generate variants with more context (audience, goal, tone)
subject score # score a specific subject line you already have
subject ab test # structure a full A/B test plan with sample size formula
subject analyze competitors # analyze subject lines from competitor emails you paste
subject by industry # get industry-specific benchmarks and top-performing patterns
subject history # show previously tested subject lines and their scores
subject save # save current session results to ~/email-subjects/
What Data to Provide
The agent works with:
No integrations required. Works entirely from your descriptions.
Workspace
Creates ~/email-subjects/ containing:
memory.md β saved audience profiles, brand voice notes, and past A/B test resultshistory/ β past testing sessions saved as markdown (session-YYYY-MM-DD.md)benchmarks.md β industry benchmark reference updated during sessionsAnalysis Framework
1. The 7 Open-Rate Predictor Signals
Each subject line is scored 0-10 on each signal; total score is out of 70:
Signal 1 β Urgency Words
Signal 2 β Personalization Tokens
Signal 3 β Question Format
Signal 4 β Number Inclusion
Signal 5 β Emoji Presence
Signal 6 β Character Length
Signal 7 β Power Words
2. Industry Benchmark Reference
| Industry | Average Open Rate | Top Quartile | |----------|------------------|--------------| | Ecommerce | 15-20% | >25% | | SaaS / Software | 20-25% | >32% | | Newsletter / Media | 25-35% | >45% | | B2B Services | 20-28% | >35% | | Nonprofit | 26-30% | >40% | | Healthcare | 22-27% | >35% |
3. Spam Trigger Detection
4. Mobile Preview Check
5. A/B Test Setup Guidance
Output Format
Every subject test run outputs:
1. 10 Scored Variants β each with total score /70, per-signal breakdown, and character count
2. Top 3 Picks β recommended for A/B testing, with rationale for each selection
3. Spam Flag Report β any variants with trigger words highlighted
4. Mobile Preview Simulation β truncated rendering at 40 and 58 characters
5. A/B Test Plan β test setup instructions with sample size recommendation
6. Preheader Suggestions β paired preheader for each top-3 variant
Rules
1. Always generate exactly 10 variants before scoring β never fewer
2. Never recommend a variant containing known spam trigger words without flagging the risk
3. Score every variant on all 7 signals β no signal may be skipped
4. Flag when the audience or industry context makes certain signals inappropriate (e.g., emoji in B2B financial services)
5. Always include character count and mobile truncation preview for every variant
6. When scoring a user-provided subject line, explain each signal score individually β not just the total
7. Save session results to ~/email-subjects/history/ when the user requests subject save
π Constraints
1. Always generate exactly 10 variants before scoring β never fewer
2. Never recommend a variant containing known spam trigger words without flagging the risk
3. Score every variant on all 7 signals β no signal may be skipped
4. Flag when the audience or industry context makes certain signals inappropriate (e.g., emoji in B2B financial services)
5. Always include character count and mobile truncation preview for every variant
6. When scoring a user-provided subject line, explain each signal score individually β not just the total
7. Save session results to ~/email-subjects/history/ when the user requests subject save