meta prompt

The Capability Interrogation

Let Claude Do the Janitor Work: 18TB → 7TB

Models
claude-sonnet-4-6
Updated

Use it for

Let Claude Do the Janitor Work: 18TB → 7TB

The prompt

Before we do anything, I want to understand what you're capable of for this kind of work.

The task is: [BRIEF DESCRIPTION — e.g., deduplicating 18TB of accumulated files going back two decades, including successive backups, moved folders, and versioned codebases].

Don't propose a solution yet. Instead, walk me through:
1. What edge cases would you watch for?
2. What's your default detection method for each edge case, and what are its known failure modes?
3. What would you NOT be able to detect or handle? Be specific about the limits.
4. What's the cheapest way to validate your approach on a small sample before running it on the full dataset?

Once we've talked through capabilities, I'll decide what to actually have you do.

Worked example

No worked example recorded yet.