ManufacturingQualitySkill

---
name: ManufacturingQualitySkill
description: >
  Generate batch-level quality reviews by combining MES exports, inspection logs, and corrective action templates.
version: 1.1.0
visibility: organization
tags:
  - manufacturing
  - quality
  - analytics
entrypoint:
  command: analyze_quality
  label: Produce batch quality digest
---

## When Claude should load this skill
- Use when diagnosing line-level defects, preparing weekly CAPA reviews, or briefing plant leadership on yield risk.
- Schedule nightly rollups with `cron: 0 2 * * *` when running inside Claude API automations.

## Core workflow
1. Load `instructions/quality-review.md` for investigation flow, KPI thresholds, and escalation criteria.
2. Execute `python scripts/quality_tracker.py` to emit `QUALITY_ALERTS` with top defects per batch.
3. Populate `templates/capa-plan.md` to assign owners, root-cause hypotheses, and containment steps.

## Linked resources
- `data/mes_batches.csv` — Production metadata including shift, line, and supervisor notes.
- `data/inspection_records.csv` — Visual inspection results with defect taxonomy and counts.
- `instructions/quality-review.md` — Troubleshooting questions, sampling guidance, and reporting cadence.
- `templates/capa-plan.md` — CAPA template for documenting corrective and preventive actions.
- `dashboards/quality-kpis.csv` — Optional KPI snapshot for trending scrap rate and first-pass yield.

import pandas as pd

batches = pd.read_csv("data/mes_batches.csv")
inspections = pd.read_csv("data/inspection_records.csv")

defects = inspections.groupby(["batch_id", "defect_type"]).size().reset_index(name="count")
merged = batches.merge(defects, on="batch_id", how="left").fillna({"count": 0})

top_defects = (
    merged.sort_values("count", ascending=False)
    .groupby("batch_id")
    .head(3)
    .to_dict("records")
)

print("QUALITY_ALERTS=", top_defects)