Open source · TypeScript · tests · 2026-05-12

Fixing multiline slash-command handling in reward parsing

I worked on a small parsing issue in ubiquity-os-marketplace/text-conversation-rewards. The problem was narrow but easy to miss: slash commands could span multiple lines, and the reward parser needed to ignore command-only content instead of treating it as meaningful comment text.

What I changed

The fix removes slash-command blocks from comment text before reward evaluation. It covers cases like /ask followed by blank lines and additional command text, while preserving useful content that appears before a command and avoiding false positives for inline slash mentions.

How I checked it

I added regression coverage around multiline commands and ran both targeted and full project checks:

• bun x prettier --check src/parser/data-purge-module.ts tests/purging.test.ts
• bun x eslint src/parser/data-purge-module.ts tests/purging.test.ts
• bun x jest tests/purging.test.ts --runInBand
• bun run build
• bun run jest:test --runInBand — 25 suites and 156 tests passed

What I learned

Parser fixes are mostly about boundaries. The implementation matters, but the test cases matter more: command-only comments, useful text before a command, and ordinary inline slash usage all need to be separated. Without those examples, it is too easy to fix one case and quietly break another.

Branch: fix/242-multiline-slash-command-purge
Compare: open upstream compare