csv quality triage

    1

    Turn messy CSV/TSV files into prioritized, evidence-backed quality tickets for reliable data analysis.

    Free

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    csv quality triage

    Example session with this skill installed

    Audit data/sales_q3_final.csv for a YoY revenue comparison. The primary key should be store_id and date. Use GBK encoding.

    • Read your context and instructions
    • Compiled the csv quality triage
    • Generated the data export

    The audit of data/sales_q3_final.csv is complete. While the file contains structural data, it is currently

    unfit for a Year-over-Year (YoY) comparison because it only contains data from July to September 2023, providing no 2022 baseline for comparison.

    Data Quality Triage Report

    Verdict: BLOCK (Score: 42/100)
    File: data/sales_q3_final.csv (9,420 rows x 12 columns)
    Identity: Encoding: GB18030 | Delimiter: , | SHA-256: 8f2a1b...

    Fix First (Critical)

    • [TYPE-02] Mixed Types in Column revenue

      • Affected: 12 rows (0.1%)
      • Evidence: Rows 150, 412, 1085 contain string values ("无", "N/A") in a numeric column.
      • Impact: Mathematical operations and YoY sums will fail or skip these records.
      • Fix: Replace non-numeric tokens with empty values or zeros in the source export.
    • [STRUCT-05] Embedded Summary Rows

      • Affected: 3 rows
      • Evidence: Rows 3140, 6280, 9420 contain "合计" (Total) labels instead of store_id.
      • Impact: These rows double-count the revenue, inflating all aggregates by 100%.
      • Fix: Filter out rows where store_id is "合计" before analysis.

    Confirm with Data Owner (Warning)

    • [GRAIN-01] Duplicate Grain Records
      • Affected: 45 rows
      • Evidence: Multiple entries for store_id 882 on 2023-08-12.
      • Question: Should these be summed together, or is one a correction of the other?

    Column Snapshot

    ColumnType MixMissing %Distinct
    revenuefloat/str0.1%8,122
    store_idint0%120
    datedate0%92

    What this report cannot tell you

    1. Whether the revenue figures themselves are factually accurate compared to bank statements.
    2. If store_id values are valid in your CRM; this only checks for internal consistency and duplicates.

    Next steps

    1. Remove the "合计" rows and re-run the export to include Q3 2022 for YoY analysis.
    2. To verify the fix, run: python3 scripts/audit.py "data/sales_q3_final_v2.csv" --key store_id,date

    sales_q3_final.csv

    CSV · data export

    Generated

    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

    Data analysis fails when the underlying CSV is broken. Mixed types, garbled encoding, and hidden totals often go unnoticed until a model crashes or a report shows impossible numbers. This skill provides a deterministic triage layer for delimited files, turning messy data into a prioritized ticket list for immediate fixing.

    What it does

    • Audit integrity by detecting encoding issues, mojibake, and delimiter mismatches before analysis begins.
    • Surface structural defects including duplicate rows, missing values, and inconsistent identifier keys.
    • Identify mixed data such as summary rows, totals, or sub-headings embedded within raw data lines.
    • Validate analysis fitness by checking if the file's time span, grain, and dimensions match the user's specific business purpose.
    • Produce fix-ready tickets that include the specific rule fired, affected row counts, and the exact action needed to repair the data.

    How it works

    1. Define intent by providing the file path and the specific purpose for the data (e.g., "monthly report" or "regression").
    2. Execute audit via a specialized script that extracts metrics, verdicts, and findings without manual file inspection.
    3. Review fitness where the agent compares the file's coverage against the user's requirements, lead-reporting any gaps.
    4. Triage findings into critical, warning, and info tickets based on their impact on downstream analysis.

    Frameworks & tools

    This skill utilizes a Python-based auditor script, scripts/audit.py, for deterministic data validation. It processes local CSV and TSV files, supporting multiple encodings like GBK/GB18030 for Chinese-locale exports.

    Why this beats prompting it yourself

    Generic prompts often miss "invisible" errors like embedded total rows or incorrect grain that look like valid data. This skill uses a formal rule engine and machine-generated metrics to ensure every count and row number is exact, preventing hallucinations common in manual data reviews.

    Use cases

    • Pre-regression audit to ensure column types are numeric and no sentinel values skew the results.
    • Export validation for files from Excel or survey tools to find garbled text or mixed date formats.
    • Comparison preparation to verify a file contains the necessary baseline periods for YoY or QoQ reporting.
    • PII hygiene check to identify and mask personal data columns before sharing analysis results.

    Known limitations

    Does not support Excel (.xlsx) files, database connections, or files over 200 MB. It is a read-only tool that identifies problems but never edits or fixes the source file itself.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for LLM Ops.

    No .xlsx support (CSV/TSV only). Max file size 200MB. Read-only; does not edit or clean data. No statistical modeling or visualization.

    How to use

    1. 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. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    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

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed today

    • Free with an Agensi account

    Software
    Python

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Listedtoday

    What's inside

    Frequently Asked Questions

    The skill uses a deterministic Python-based engine to scan for 'silent' failures like embedded totals, mixed encodings (including GBK/GB18030), and grain inconsistencies that standard LLM prompts often hallucinate or overlook.

    This skill currently supports local CSV and TSV files up to 200MB, but it does not support Excel (.xlsx) formats or direct database connections.

    You will receive the core logic engine, the specialized `audit.py` script for local execution, and the integration layer that allows your AI agent to interpret and prioritize the audit findings.

    No, this is a read-only diagnostic tool; it generates a prioritized 'ticket list' of specific errors and row numbers so you can fix the source file safely without risking automated data corruption.

    The skill is designed to work with any AI agent capable of executing Python scripts and processing JSON-based diagnostic reports, making it ideal for LLM-ops workflows.

    By defining your 'intent' (e.g., 'YoY reporting') during setup, the skill validates if the file actually contains the necessary date ranges and dimensions required for that specific business goal.

    New in LLM Ops

    Free