fix(extraction): change cases abstract to descriptive format (Issue #640)

[jlin53882]

Summary

Fix Issue #640: Smart-extractor skips [cases] category due to misleading BigInt prompt guidance

Problem

The abstract in cases category prompt was written in imperative format:

abstract: "LanceDB BigInt error -> Use Number() coercion before arithmetic"

This made the LLM interpret it as "skip memories about this" rather than "this is what the case is about", causing:

• stats.created = 0
• stats.merged = 0
• stats.skipped > 0
• Triggers "falling back to regex capture"

Solution

Changed abstract from imperative to descriptive format:

abstract: "LanceDB BigInt numeric handling issue"

This describes what the case IS, not what to DO with it.

Changes

• src/extraction-prompts.ts: Updated cases category abstract
• test/issue-640-bigint-prompt.test.mjs: Added unit test

Test Results

=== Issue #640: BigInt Prompt Format Test ===
[Old format (buggy) - should be detected as misleading]
  Misleading: true (expected: true) ✅ PASS

[New format (fixed) - should NOT be misleading]  
  Misleading: false (expected: false) ✅ PASS
----------------------------------------
Total: 2 passed, 0 failed
----------------------------------------

Related

• Issue [BUG] Smart-extractor skips [cases] category due to misleading BigInt prompt guidance #640

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.