- Home
- Skills
- Data & Databases
- CSV Import Validator: Exact line and column of bad rows
More screenshots
Works with the AI Agents you already use
CSV Import Validator: Exact line and column of bad rows
The import fails halfway, or worse, succeeds with bad data. Check a CSV against your column rules and get the exact row, column and code for every bad value, duplicate key and short row.
$12
CSV Import Validator: Exact line and column of bad rows
Example session with this skill installed
Check references/failing.csv with references/example-schema.json before import. Return Markdown and prioritize fixes without changing the file.
- Read your context and instructions
- Compiled the csv import validator
- Generated the data export
Captured stdout from python3 -I -B scripts/csv_validate.py references/failing.csv --schema references/example-schema.json --format markdown (exit 1):
CSV validation
Valid: false
Checked rows: 4; errors: 10
UTF-8 BOM: false; complete: true
| Row | Column | Field | Code | Message |
|---|---|---|---|---|
| 3 | 2 | count | min | Value violates inclusive min bound. |
| 3 | 3 | amount | regex | Value does not fully match the schema regex. |
| 3 | 4 | day | type | Calendar date does not exist. |
| 3 | 5 | status | enum | Value is not in the allowed enum list. |
| 3 | 1 | record_id | duplicate_key | Duplicate key on columns: record_id. First seen at row 2. |
| 4 | 1 | record_id | required | A nonempty value is required. |
| 4 | 2 | count | max | Value violates inclusive max bound. |
| 4 | 3 | amount | type | Expected a plain decimal; no exponent, spaces or leading zeros. |
| 4 | 5 | status | duplicate_key | Duplicate key on columns: day, status. First seen at row 2. |
| 5 | — | — | columns | Expected 6 columns; found 2. |
FAIL: hold the import. All 4 records were checked; 10 findings affect 3 records. Restore the missing columns on line 5 first, resolve duplicate keys against line 2 with the data owner, then correct required, numeric, enum and date violations. Rerun the same contract. No data was changed.
failing.csv
CSV · data export
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
The problem Someone hands you a CSV export and you need to load it. The header looks right, but that tells you nothing about a February 30th date, an amount like 1e3, or the same customer ID showing up twice. You'd rather find those before the import fails halfway, or worse, succeeds.
See it before you buy: real input, real output Command (from the package folder, exit code 1, 0.025 s):
python3 -I -B scripts/csv_validate.py references/failing.csv --schema references/example-schema.json --format markdown
Before: references/failing.csv
record_id,count,amount,day,status,label
EXAMPLE-01,2,12.50,2024-02-29,test,Synthetic
EXAMPLE-01,-1,1.234,2026-02-30,unknown,
,101,NaN,2024-02-29,test,Another
EXAMPLE-04,3
After: lines copied from the script's output (excerpt):
Valid: false
Checked rows: 4; errors: 10
| 3 | 4 | day | type | Calendar date does not exist. |
| 3 | 1 | record_id | duplicate_key | Duplicate key on columns: record_id. First seen at row 2. |
| 4 | 3 | amount | type | Expected a plain decimal; no exponent, spaces or leading zeros. |
| 4 | 5 | status | duplicate_key | Duplicate key on columns: day, status. First seen at row 2. |
What the numbers show
- Ran on 3 public files (763 records: country codes and two S&P 500 lists). The originals produced 0 problems; copies with 6 planted mistakes each produced 6.
- All 18 reports were checked by hand: every planted mistake at the right line, nothing extra. Each run took under 0.05 seconds.
- Honest limit: a small test with mistakes I chose, so it doesn't prove accuracy on your files.
What you get
- One Python script plus instructions for your agent. The agent reviews your schema with you, runs the check and tells you which fixes to make first.
- You describe each column in a small JSON file: its type (int, decimal, date, enum or str), whether it's required, min/max bounds and an optional regex. Headers must match in the exact order.
- Duplicate checks on one column or a combination of columns, such as day + status. Each duplicate points back to the line where the value first appeared. 1.0 and 1.00 count as the same decimal key.
- A JSON or Markdown report that gives the file line and column for each problem, using 15 named problem codes.
- Fixed limits: 1 MiB per file, 5,000 records, 100 columns. Reports list up to 100 problems, and the JSON still gives the full count and says whether the whole file was read.
- Schema docs, example files, 21 test cases and 4 saved sample reports.
Measured on real data (full detail) I ran it on three public files: a country-code list and two S&P 500 company lists. I wrote a schema for each, then made a copy of each file with 6 deliberate mistakes (blank required values, a bad date, a repeated key, a short row and so on).
| File | Records | Problems in original | Problems in broken copy |
|---|---|---|---|
| country | 249 | 0 | 6 |
| constituents | 503 | 0 | 6 |
| sector-counts | 11 | 0 | 6 |
All 763 original records passed. The broken copies produced 18 reports, and I checked each one by hand: all 18 were the mistakes I'd planted, at the right line, with nothing extra. Each run took under 0.05 seconds, including Python startup. That's a small test with mistakes I chose, so it doesn't prove accuracy on your files. references/benchmarks.md has the snapshot versions, schemas, commands and every finding.
Known limitations
- It doesn't import, repair or guess a schema. You write the rules.
- It doesn't check your database's constraints or links between files.
- Input must be UTF-8 (a BOM is fine). Regex checks have no timeout, so look over your patterns before running.
- A file over the limits always fails. If you split it into batches, duplicates are only caught within each batch.
Who it's for Developers and data people who know what their import expects and want the same check on their laptop and in CI.
Requires Python 3.8 or newer. No packages to install. You provide the CSV and a JSON schema file. The script runs on its own and doesn't need the other package files next to it.
Runs locally with no network, no telemetry and no file writes unless you redirect output.
Check it yourself
21 test cases pass, and 4 saved sample reports match a fresh run byte for byte. I also reran the 3 public files and their broken copies. To try it yourself from the package folder, run python3 -I -B scripts/csv_validate.py references/passing.csv --schema references/example-schema.json (exits 0), then swap in references/failing.csv (exits 1). The test list is in references/checks.json and the measurement steps are in references/benchmarks.md.
Price and guarantee $12 one time, yours to use with any agent that can run Python. Covered by Agensi's money-back guarantee if it doesn't do what this page promises.
1 MiB and 5,000 record limit. No auto-repairs, foreign key checks, or spreadsheet formula execution. Requires exact YYYY-MM-DD date formats.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 15 days ago