nicholashidalgo email-writer
Purpose: Convert a brief or messy email draft into a direct, natural-sounding professional email free of AI tells.
Input Schema
| Field | Type | Required | Notes |
|---|---|---|---|
subject_line |
string | yes | Proposed subject or topic |
intent |
string | yes | What the email needs to accomplish |
key_facts |
string[] | yes | Facts, names, dates that must be preserved |
tone |
string | yes | One of: direct, formal, concise, warm-professional |
raw_draft |
string | no | Rough draft to clean up |
{
"subject_line": "API deprecation timeline",
"intent": "Notify the payments team that v1 API endpoints are deprecated and cutover is required by March 31.",
"key_facts": ["March 31 deadline", "v1 API", "payments team", "docs at /api/migration"],
"tone": "direct",
"raw_draft": "I hope this email finds you well! I wanted to reach out and touch base regarding the important and crucial matter of our API deprecation timeline going forward."
}
Output Schema
{
"subject": "v1 API deprecation - cutover required by March 31",
"body": "The v1 API endpoints will be shut off March 31. If your team is still using them, migrate before then - the guide is at /api/migration. Reply here or ping me in Slack if you hit anything unexpected.",
"patterns_removed": [7, 19, 22],
"word_count": 52
}
Prompt Flow
Pass 1: Apply core email rewrite. Remove: chatbot openers (P19), sycophancy (P21), em dashes (P13), AI vocab (P7), filler (P22), negative parallelisms (P9), excessive hedging (P23), copula avoidance (P8). Preserve key_facts.
Pass 2 Audit: "What makes this email so obviously AI-generated?" - list remaining tells.
Pass 2 Final: Rewrite removing remaining tells. Keep under 150 words unless facts require more.
Examples
Short - API notification
Before: "I hope this email finds you well! I wanted to reach out regarding the crucial API deprecation timeline, which stands as a pivotal moment in our infrastructure evolution." After: "The v1 API endpoints shut down March 31. Migrate before then - guide at /api/migration."
Medium - Status update
Before: "Just wanted to circle back and provide an update. At this point in time, we have completed the initial phase of the project." After: "Quick status: phase 1 is done. Phase 2 (auth integration) starts Monday. We're on track for the April 15 launch."
Long - Escalation with context
Before: "Great question regarding the outage. Additionally, industry observers have noted that distributed systems face challenges typical of complex environments. The future looks bright." After: "The March 14 outage (9:41-11:23 PM PT) was a misconfigured load balancer after the 11 PM deploy. 4,300 requests timed out. Fix: config rollback. Post-mortem at [link]. I've added a deploy checklist item to catch this going forward."
Unit Tests
# tests/skills/test_email_writer.py
import pytest
from ai_pattern_scrubber import detect_patterns
SHORT_FINAL = "The v1 API endpoints shut down March 31. Migrate before then - guide at /api/migration."
MEDIUM_FINAL = "Quick status: phase 1 is done. Phase 2 (auth integration) starts Monday. On track for April 15."
LONG_FINAL = (
"The March 14 outage (9:41\u201311:23 PM PT) was a misconfigured load balancer. "
"4,300 requests timed out. Fix: config rollback. Post-mortem at [link]."
)
def high_severity_hits(text):
return [h for h in detect_patterns(text) if h.severity == "high"]
def test_email_short_no_high_severity_flags():
assert high_severity_hits(SHORT_FINAL) == []
def test_email_medium_no_high_severity_flags():
assert high_severity_hits(MEDIUM_FINAL) == []
def test_email_long_no_high_severity_flags():
assert high_severity_hits(LONG_FINAL) == []