Resume Bullet Rewriter

resume-bullet-rewriter

Purpose: Targeted single-bullet rewriter. Accepts one weak bullet and returns a strengthened version plus a metric-gap warning if no number is available.

Input Schema

{
  "bullet": "Responsible for improving the checkout experience and working with design.",
  "context": "Senior Product Manager, e-commerce startup",
  "available_metrics": ["checkout conversion up 12%", "reduced 3-step flow to 1-step"],
  "preserve_facts": ["12%", "1-step"]
}

Output Schema

{
  "original": "...",
  "revised": "...",
  "changes": ["removed 'responsible for'", "added metric 12%", "started with action verb 'Redesigned'"],
  "metric_warning": null
}

Prompt Flow

Pass 1: Rewrite starting with action verb. Inject best available metric from available_metrics. Remove all banned phrases. Preserve preserve_facts.

Pass 2: If no metric available, output revised bullet + metric_warning: "No metric available - add a number before publishing." Never fabricate a metric.

Examples

Short

Before: "Responsible for improving the checkout experience." After: "Redesigned checkout from a 3-step to 1-step flow, lifting conversion by 12%."

Medium

Before: "Helped to streamline the onboarding process and assisted with implementation." After: "Wrote the engineer onboarding runbook adopted across 3 teams; ramp time dropped from 6 weeks to 3."

Long (with context)

Before: "Results-driven PM passionate about delivering innovative solutions that leverage data to drive business outcomes at scale." After: "Shipped 6 product features in Q2 2024 driven by customer cohort analysis; 4 of 6 exceeded their 90-day retention targets."

Unit Tests

# tests/skills/test_resume_bullet_rewriter.py
from resume_banned import flag_banned_phrases
from ai_pattern_scrubber import detect_patterns
import re

REVISED = "Redesigned checkout from a 3-step to 1-step flow, lifting conversion by 12%."
METRIC_PATTERN = re.compile(r'\d+[%xKMB]?|\$[\d,]+|\d+\s*(weeks?|days?|steps?)')

def test_revised_no_banned_phrases():
    assert flag_banned_phrases(REVISED) == []

def test_revised_contains_metric():
    assert METRIC_PATTERN.search(REVISED), "Revised bullet must contain a metric"

def test_revised_no_high_severity():
    assert not [h for h in detect_patterns(REVISED) if h.severity == "high"]