ccr-summary

Files (3)

• SKILL.md
• example.md
• requirements.md

SKILL.md (first lines)

# CCR Summary (40 CFR § 141.156)

Generate the **opening summary** for a CCR by reading the full report PDF and mapping findings to [40 CFR § 141.156](https://www.law.cornell.edu/cfr/text/40/141.156). The summary is **not** a rewrite of the whole report. It states what consumers need first.

**Scope note:** This workflow **excludes § 141.156(c)(2)** (limited English proficiency / translated reports). Only include that block if the user explicitly asks and the system is subject to § 141.153(h)(3).

**Local references** (in this repo): `40 CFR § 141.153…html`, `40 CFR § 141.155…html`, `40 CFR § 141.156…html`.

For the full regulatory checklist and PDF search patterns, read [requirements.md](requirements.md). For a worked small-system example, read [example.md](example.md).

---

## Heading rule (do not skip)

**All subsection headings in summary output MUST be h5 - five hash marks (`#####`) in markdown.**

| Correct | Wrong |
|-----------|---------|
| `##### Your water at a glance` | `## Your water at a glance` |
| `##### How to get more information` | `### How to get more information` |
| `##### Paper copy of this report` | `#### Paper copy of this report` |

- **Opening paragraphs:** no heading (the printed report supplies the title).
- **Subsections below the opening:** `#####` only - never `#`, `##`, `###`, or `####`.
- **§ 141.156(f) sharing block:** omit the heading if the layout supplies it; if you include one, it must still be `##### Please share this report`.

This matches the CCR layout template. Using `##` or other levels will break formatting.

---

## Workflow

Copy this checklist and mark items as you go:

```
CCR summary progress:
- [ ] Step 1: Extract full PDF text
- [ ] Step 2: Identify system metadata and delivery context
- [ ] Step 3: Scan for violations/compliance (required sources)
- [ ] Step 4: Extract contact information
- [ ] Step 5: Evaluate conditional § 141.156(c) and (e) items
- [ ] Step 6: Draft summary using template (**subsection headings = `#####` / h5 only**)
- [ ] Step 7: Compliance self-check (**verify h5 headings before sending**)
```

### Step 1: Extract full PDF text

Do not rely on partial reads. Extract **every page** to a UTF-8 text file (avoids Windows console encoding errors):

```python
from pypdf import PdfReader

path = r"<absolute-path-to-ccr.pdf>"
out = r"<absolute-path-to-extract.txt>"
reader = PdfReader(path)
with open(out, "w", encoding="utf-8") as f:
    for i, page in enumerate(reader.pages):
        f.write(f"--- PAGE {i+1} ---\n")
        f.write(page.extract_text() or "")
        f.write("\n\n")
```

If `pypdf` is missing: `pip install pypdf`. If extraction is empty or garbled, tell the user the PDF may be scanned - OCR will be needed before summarizing.

### Step 2: System metadata and delivery context

Record from the report (usually cover, footer, or “System Contact”):

| Field | Where to look |
|-------|----------------|
| System name | Title / header |
| PWSID | Header or contact block |
| Report year / monitoring period | Title (“For The Year Of …”) |
| Population served | Sometimes in header or certification (not always in PDF) |
| Delivery method | Cover “online version,” QR code, website URL, mail/email notices |

**Delivery drives conditionals:**

- **Paper mail/hand delivery only** → § 141.156(c)(1) usually **not** needed.
- **Website link, email link, or other electronic delivery** (§ 141.155(a)(1)(ii)-(iv)) → summary **must** include directions to request a **paper copy** per § 141.155(a)(2). Use the system’s actual request process if stated; otherwise use the regulatory minimum (contact phone/email/mail).

**Biannual / 6-month update (§ 141.156(e)):** Required only if the system serves **≥ 10,000 persons** and distributes twice per year (§ 141.155(j)(2)). Small systems like mobile home parks typically **skip (e)** unless the report itself states biannual delivery.

### Step 3: Violations and compliance summary (§ 141.156(b)(1))

Summarize **only what appears in the report** for the monitoring period covered. Source sections in the **full report** (not only the summary you are writing):

| CFR source | What to find in the PDF |
|------------|-------------------------|
| **§ 141.153(d)(6)** | Contaminant tables: `Violation` = Y/Yes; values over MCL/MRDL/TT; narrative “violation,” “exceeded,” “failed to meet.” Include length, health effects language, and corrective actions if provided. |
| **§ 141.153(d)(8)** | Lead **action level** exceedance: 90th percentile or `Samples > AL` > 0; lead exceedance explanation; consumer steps to reduce exposure; corrective actions. |
| **§ 141.153(f)** | NPDWR compliance violations **other than** contaminant-level exceedances already in (d)(6): monitoring/reporting failures, filtration/disinfection (subpart H), lead/copper **rule** violations (failure to complete required actions - not the same as AL exceedance in (d)(8)), subpart K, recordkeeping, §§ 141.40/141.41 monitoring, variance/exemption/order violations. |
| **§ 141.153(h)(6)** | Groundwater Rule (subpart S): uncorrected **significant deficiency**, fecal indicator-positive source, state-approved correction plan - often a dedicated section. |
| **§ 141.153(h)(7)** | Revised Total Coliform Rule (subpart Y): Level 1/2 assessments, E. coli MCL, missed assessments/corrective actions - often a dedicated section. |

**If no violations in any category:** State clearly that the report shows **no violations** of drinking water standards for the period covered (still required content for (b)(1)).

**Do not treat as violations:**

- Educational arsenic/lead text when MCL/AL are met
- Monitoring **waivers** (e.g., SOC, asbestos) unless the report says a violation occurred
- Detected contaminants **below** MCL/AL with `Violation` = N

**Testing waivers (optional one-sentence summary):** If the report has a “Testing Waivers” (or similar) section and the system had **no violations**, you may add **one short plain-language sentence** in **Your water at a glance** so readers know skipped tests are not problems. **Do not** name EPA, the state, primacy agency, or any regulator as who granted the waiver - the report may say “EPA,” “State,” or “ADEC,” but the summary must stay **agency-neutral**.

| Avoid | Use instead |
|-------|-------------|
| “when EPA allowed us to skip tests” | “we were not required to run some tests” |
| “EPA granted a waiver” | “some tests do not apply to our system” |
| “ADEC/the State approved…” | (omit agency name entirely) |

**Example (plain language):** “We also did not run some tests (for example, asbestos and certain chemicals) because they do not apply to our system. Skipping those tests is **not** a violation.”

See [requirements.md](requirements.md) for subsection detail and search keywords.

### Step 4: Contact information (§ 141.156(b)(2) / § 141.153(h)(2))

Include **telephone** (required) for owner, operator, or designee. Also include name, email, and address when present. Prefer the block labeled “System Contact,” “Operator Contact,” or end-of-report contact.

From KroniK907/agent-config, MIT.

More from KroniK907/agent-config