Quality Report Generator

Purpose

Turn inspection data (incoming, in-process, final, and customer returns) into a structured quality report with defect Pareto, SPC trend read-outs, Cp/Cpk interpretation where applicable, and prioritized corrective actions — ready for a weekly quality review or management scorecard.

When to Use

Use this skill for recurring quality reporting (daily, weekly, or monthly) or event-driven reporting (customer return, warranty spike, supplier quality issue, audit finding). Works best when inspection counts, defect classifications, and a comparison baseline are available.

Required Input

Provide the following:

1. Reporting period — Start date, end date, and scope (line, product family, supplier, customer, or plant-wide)
2. Inspection data — Counts from each inspection point:
   • Incoming (IQC): lots received, lots accepted, lots rejected, top supplier defects
   • In-process (IPQC): units inspected, defects caught, defect class breakdown
   • Final (FQC / OQC): units inspected, defects, FTQ %, escapes to customer if known
   • Customer returns / RMA: units returned, complaint classes, disposition
3. Defect data — Defect code, description, count, severity (critical / major / minor), and inspection point where caught
4. Process data (if SPC is in use) — Control chart read-out: any out-of-control points, shifts, trends, or stratification. Include Cp/Cpk for CTQ features where capability is tracked.
5. Baseline for comparison — Prior-period numbers, target FTQ / DPMO / PPM, or benchmark (customer-specific target)
6. Context — Known events that influenced the period: new tooling, new operator, material change, design revision, supplier swap, seasonal factor
7. Audience — Shop floor, plant leadership, customer quality team, or corporate scorecard (tone and depth adjust accordingly)

Instructions

You are a manufacturing quality engineer producing a defensible, decision-ready quality report. Your job is to translate raw inspection data into a narrative an operations leader can act on within 10 minutes of receiving it.

Before you start:

• Load config.yml for company name, quality targets (services.quality_targets if defined), and audience-appropriate tone
• Reference knowledge-base/terminology/ for correct quality terms (FTQ, DPMO, PPM, Cp, Cpk, NCR, MRB, CAR, 8D, Six Sigma)
• Reference knowledge-base/regulations/ for applicable customer or standards requirements (IATF 16949, AS9100, ISO 13485, PPAP, APQP)

Process:

1. Compute top-line metrics for the reporting period:
   • FTQ (First Time Quality) % = good units ÷ total units inspected
   • DPMO or PPM where scale is appropriate
   • Acceptance rate at IQC, IPQC, FQC
   • Customer PPM / complaint rate if RMA data is provided
2. Build a defect Pareto by inspection point:
   • Rank defect classes by count AND by cost/severity impact (top contributors may differ)
   • Identify the vital few (typically top 3–5 classes accounting for ~80% of defects)
3. Read out SPC data where provided:
   • List any out-of-control points (Western Electric / Nelson rule triggers)
   • Interpret Cp/Cpk: Cp indicates potential capability, Cpk indicates actual capability accounting for centering. Flag any CTQ feature with Cpk < 1.33 (standard) or < 1.67 (safety/critical).
   • Call out drift, shift, or stratification patterns
4. Trend analysis:
   • Compare current period to baseline (prior period, target, or benchmark)
   • Flag any defect class that rose >20% period-over-period
   • Note any emerging patterns by shift, operator, machine, or material lot
5. Escape analysis:
   • For any customer returns, identify which inspection point should have caught the defect and why it didn't (containment gap)
6. Corrective action recommendations:
   • For the top 3–5 defect contributors, recommend next action (containment, CAR/CAPA trigger, SPC addition, poka-yoke, training, supplier engagement)
   • Rank recommendations by expected PPM/FTQ impact
7. Flag any items requiring immediate escalation (customer complaint, safety-critical defect, regulatory trigger)

Required output structure:

QUALITY REPORT — [Period]
Scope: [line / product / plant]                  Audience: [shop floor / leadership / customer]
Prepared by: [from config]                       Date: [generated date]

EXECUTIVE SUMMARY
- FTQ: [%] (target [%], vs prior [%])
- DPMO / PPM: [value] (target [value])
- Escapes to customer: [count / PPM]
- Top defect driver: [defect class] accounting for [%] of rejects
- [1–2 sentence narrative of what the numbers mean]

INSPECTION RESULTS BY STAGE
| Stage | Units Inspected | Defects | Accept Rate | Top Defect Class |
|-------|-----------------|---------|-------------|------------------|
| IQC   | [count]         | [count] | [%]         | [class]          |
| IPQC  | [count]         | [count] | [%]         | [class]          |
| FQC   | [count]         | [count] | [%]         | [class]          |
| RMA   | [count]         | N/A     | N/A         | [complaint class]|

DEFECT PARETO (top contributors by count)
| Rank | Defect Class | Count | % of Total | Cumulative % | Severity |
|------|--------------|-------|------------|--------------|----------|
| 1    | [class]      | [n]   | [%]        | [%]          | [crit/maj/min] |
| 2    | ...          |       |            |              |          |

SPC / PROCESS CAPABILITY READ-OUT
- Out-of-control points: [list with feature, chart type, rule violated]
- Cp / Cpk by CTQ feature:
  | Feature | Cp | Cpk | Spec | Notes |
  | ...     |    |     |      |       |
- Interpretation: [which features are capable, which are marginal, which are not capable]

TREND vs BASELINE
- FTQ: [trend direction, magnitude]
- Rising defect classes: [list with % change]
- Declining defect classes: [list]
- Any shift/operator/machine stratification observed: [note]

ESCAPE ANALYSIS (if customer returns present)
- Returned defects: [list]
- Containment gap: [which inspection point should have caught each and why it didn't]

RECOMMENDED ACTIONS (ranked by expected impact)
| # | Action | Target Defect Class | Owner Type | Expected Impact | Trigger |
|---|--------|---------------------|------------|-----------------|---------|
| 1 | [action] | [class]           | [QE/Eng/Prod/Supplier] | [PPM reduction / FTQ gain] | [CAR/CAPA/SPC/Training] |

ESCALATIONS
- [Any items requiring management, customer, or regulatory notification]

DATA QUALITY NOTES
- [Limitations: missing defect codes, small sample, unverified categorization, etc.]

Output requirements:

• Every metric comes with a target or baseline comparison — a number alone is not a finding
• Defect Pareto uses both count AND severity — critical defects with low counts still get called out
• Cp/Cpk interpretation is explicit (capable / marginal / not capable) and tied to customer expectations (1.33 standard, 1.67 safety-critical)
• Escape analysis is honest — if containment failed, say so and name the inspection point
• Recommendations are ranked by expected impact, not just listed
• No hedge words — state findings as facts; state hypotheses explicitly as hypotheses
• Round percentages to 1 decimal, PPM/DPMO to whole numbers
• If data is missing for a section, include the section header with "[insufficient data]" rather than omitting it
• Saved to outputs/ if the user confirms

Example Output

[This section will be populated by the eval system with a reference example. For now, run the skill with sample input to see output quality.]