handling-customer-data

You must NEVER see customer data directly. When working with data that may contain customer information (JIDs, emails, phone numbers, names, account IDs, etc.), filter through shell tools to answer questions about the data without the data itself appearing in your context.

Core principle

Answer questions ABOUT data, not WITH data. The user may see data directly; you may not.

Approved patterns

Counting matches

redis-cli KEYS "pattern-*" | wc -l
grep -c "pattern" file.txt

Boolean checks

redis-cli KEYS "pattern-*" | grep -qE "[A-Z]" && echo TRUE || echo FALSE

Saving for user review

redis-cli KEYS "pattern-*" | grep -E "[A-Z]" > results.txt
# Report: "Saved N results to results.txt"

Aggregations

redis-cli KEYS "prefix-*" | wc -l
cat file.txt | cut -d',' -f2 | sort | uniq -c | sort -rn | head -5
# Report: "Top 5 categories by count: [counts only, not values]"

Prohibited patterns

Never do these:

# DON'T: Print matching data
redis-cli KEYS "pattern-*" | grep "something"

# DON'T: Show file contents with customer data
cat customer_jids.txt

# DON'T: Display query results
redis-cli GET "jmp_customer_jid-12345"

# DON'T: Print filtered results
grep "error" /var/log/app.log  # if log contains customer data

Decision flow

1. Does this data source contain customer information?

- Redis keys/values with JIDs, emails, phones → Yes - Application logs → Likely yes - Database tables with customer columns → Yes - Configuration files → Usually no

1. What question am I answering?

- "How many?" → Use wc -l, grep -c - "Does X exist?" → Use grep -q && echo TRUE || echo FALSE - "Which ones match?" → Save to file, report count only - "What are the values?" → Stop. Ask user to look directly.

1. Will the output appear in my response?

- If yes and contains customer data → Redirect to file instead - If yes and only contains counts/booleans → Acceptable

Example session

User: "Are there any uppercase JIDs in Redis?"

Good:

redis-cli KEYS "jmp_customer_id-*" | grep -qE "[A-Z]" && echo TRUE || echo FALSE

Response: "TRUE - there are uppercase characters in some JID keys."

User: "How many?"

Good:

redis-cli KEYS "jmp_customer_id-*" | grep -cE "[A-Z]"

Response: "52 keys contain uppercase characters."

User: "Which ones?"

Good:

redis-cli KEYS "jmp_customer_id-*" | grep -E "[A-Z]" > uppercase_jids.txt

Response: "Saved to uppercase_jids.txt. You can review the file directly."

Bad (never do this):

redis-cli KEYS "jmp_customer_id-*" | grep -E "[A-Z]"

This would print customer JIDs into the response.

When in doubt

If you're uncertain whether output contains customer data, assume it does. Redirect to a file and let the user review.