anonymize-prompt

Anonymize Prompt

Generate a privacy-safe, anonymized prompt that the user can paste into external search engines and AI chatbots to get help — without exposing ANY proprietary details.

Inputs

1. Conversation context — the current conversation where the problem is being discussed
2. User's argument (optional) — a plain-text description of the problem passed as the skill argument

Workflow

Step 1: Understand the Problem

From the conversation context and user's argument, identify:

• What is the core technical problem?
• What technologies/frameworks are involved? (these are public knowledge — keep them)
• What error messages or symptoms are observed?
• What has been tried so far?

Step 2: Anonymize Everything Proprietary

Apply ALL of the following replacements systematically:

Category	Example Before	Example After
Company/product names	AcmeCorp, my-service	"our application", "the project"
Class names	OrderManager, PaymentTracker	DataManager, ProcessTracker
Variable/method names	buildOrderInstance()	buildDataStructure()
Package paths	com.acme.app.apiImpl	com.example.app.impl
Internal URLs	localhost:8080/internal/api/v0	localhost:8080/app/api/v1
Database/table names	acme_metadata, order_definitions	app_metadata, entity_definitions
Internal jargon	proprietary domain terms	describe functionally in plain language
File paths	/src/main/java/com/acme/...	/src/main/java/com/example/...
Config keys	acme.security.*	app.security.*
Ticket IDs	PROJ-1931	remove entirely
Branch names	PROJ-1931-add-offline-access	remove entirely
Colleague names	any person names	remove entirely
Docker/service names	my-api, acme-web	"the api service", "the web service"

Step 3: Preserve Technical Accuracy

Keep ALL of the following (they are public knowledge):

• Framework names: Spring, Spring Security, Hibernate, React, Redux, Webpack, Maven, etc.
• Library versions: Spring 5.x, Java 14, PostgreSQL 12, etc.
• Standard patterns: OAuth2/OIDC, REST, JDBC, ACL, etc.
• Error messages (but scrub any proprietary class/package names within them)
• Stack trace structure (but anonymize class names)
• Configuration property formats (but anonymize the property names)

Step 4: Problem-Only Output

IMPORTANT: Only describe the problem — never include solutions, fixes, or workarounds. The goal is to get unbiased answers from external engines. Including a fix will bias the response toward confirming that fix rather than exploring all options.

• Describe the symptoms and constraints clearly
• End with an open-ended question like "What are the best ways to solve this?"
• Do NOT suggest approaches, patterns, or specific technologies as answers

Produce the anonymized prompt in this format:

## Context
[1-2 sentences: what the app is, what tech stack, what area you're working in — all anonymized]

## Problem
[Clear description of the technical issue — specific enough to be searchable]

## What I've Tried
[List of approaches attempted, if any — anonymized]

## Error / Symptoms
[Any error messages, stack traces, or unexpected behavior — anonymized]

## Question
[Open-ended question asking for solutions — do NOT suggest any answers]

Step 5: Self-Review Checklist

Before outputting, verify:

• No company or product names remain
• No internal class/method/variable names remain
• No internal file paths remain
• No proprietary package names remain
• No ticket IDs or branch names remain
• No person names remain
• No internal URLs remain
• The problem is still technically accurate and specific
• Someone reading this would have NO idea which company or project this is from

Output

Present the anonymized prompt inside a fenced code block so the user can easily copy-paste it.

Add a brief note above the code block: "Here's your anonymized prompt — safe to paste into external search engines and AI chatbots:"

If anything couldn't be fully anonymized (e.g., a very unique architectural pattern), flag it with a warning so the user can review before pasting.