Investigate Data

What it does

When your agent uses it

• The user reports a bug in autocomplete results
• Duplicate entries
• Wrong city/street/postcode results
• Asks to check data for a specific country

Example prompts

• “check the data”
• “investigate”
• “query S3”
• “/investigate-data”

Requirements

• Pre-approved tools (allowed-tools): Read, Grep, Glob, mcp__motherduck__execute_query

Workflow steps

7 steps, taken from the first numbered list in SKILL.md.

1. 1Understand the issue: What country, what index, what unexpected behavior?
2. 2Check the schema first if unsure about columns
3. 3Query the relevant index using MotherDuck MCP (mcpmotherduckexecute_query)
4. 4Diagnose the root cause using these patterns
5. 5Simulate the fix by running the corrected query and verifying output.
6. 6Cross-validate by checking 2-3 other countries for the same pattern.
7. 7Report with

What it can do on your machine

Read from SKILL.md and the folder at commit d16cf53. It shows what the files ask for, not the result of running them.

• Tool permissions

  Pre-approves these tools, so the agent can use them without asking each time:

  • Read
  • Grep
  • Glob
  • mcp__motherduck__execute_query

  From allowed-tools in the SKILL.md frontmatter.

• Runs code

  No scripts in the folder and no shell commands in SKILL.md (its code samples are sql).

  From the folder's file list and the shell code blocks in SKILL.md.

• Network

  Hosts in commands or code, which the agent is likely to contact:

  • s3.us-west-2.amazonaws.com

  From URLs in SKILL.md, links to its own repository left out.

• Credentials

  Names no API keys, tokens, secrets or passwords.

  From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Investigate Data loads about 935 tokens when it runs. Until then it costs about 132 tokens; SKILL.md has 260 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it

~132

When it runs · the whole SKILL.md, loaded when a task matches

~935

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from walkthru-earth/geocoding-playground at commit d16cf53, republished under its CC-BY-4.0 licence (© walkthru-earth). 260 words, ~935 tokens.

name

investigate-data

description

Investigates geocoder data quality issues by querying live S3 parquet files via MotherDuck MCP. Use this skill when the user reports a bug in autocomplete results, duplicate entries, missing data, wrong city/street/postcode results, or asks to check data for a specific country. Also use when the user says 'check the data', 'investigate', 'query S3', 'what does the index look like', 'how many streets/cities/postcodes', or 'verify the data'. Do NOT use for frontend UI bugs, build issues, or code refactoring.

allowed-tools

Read, Grep, Glob, mcp__motherduck__execute_query

https://s3.us-west-2.amazonaws.com/us-west-2.opendata.source.coop/walkthru-earth/indices/addresses-index/v4/release=<RELEASE>/

-- number_index: should have many small row groups + bloom filters
SELECT num_rows, num_row_groups, file_size_bytes
FROM parquet_file_metadata('...BASE_URL.../number_index/country=NL/data_0.parquet');

-- Bloom filters present?
SELECT path_in_schema, bloom_filter_offset, bloom_filter_length
FROM parquet_metadata('...BASE_URL.../number_index/country=NL/data_0.parquet')
WHERE path_in_schema = 'street_lower' LIMIT 5;

-- Geocoder tile bloom filters on street?
SELECT path_in_schema, bloom_filter_offset
FROM parquet_metadata('...BASE_URL.../geocoder/country=NL/h3_res4=841969dffffffff/bucket=_/data_0.parquet')
WHERE path_in_schema = 'street' LIMIT 3;

© walkthru-earth, CC-BY-4.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in .claude/skills/investigate-data of walkthru-earth/geocoding-playground.

Open the folder on GitHubat commit d16cf53