Vouch Request — Fix issue #3300 (superjson inline deserialization bug) #3347

Why do you want to contribute?

I've been contributing to open source AI agent and developer tooling repos for a while, and Trigger.dev's durable execution model is something I find genuinely interesting — especially around how task outputs are handled across the inline vs blob paths.

I'd like to start with issue #3300 (bug: runs.retrieve() returns null for undefined fields on small inline outputs). I've already dug into the root cause: the inline output path skips superjson deserialization (only the .json portion is used, not .meta), while the presigned/blob path correctly calls superjson.deserialize(). The fix is scoped to the deserialization utility in packages/core/src/v3/, with edge cases around plain JSON outputs, null returns, and the outputIsUndefined flag.

I'm ready to open a PR once vouched.

Prior contributions or relevant experience

I've been contributing to agent infrastructure and AI tooling repos:

CUA (Computer-Use Agents framework)

• Fix Omniparser historical message conversion using per-screenshot mappings trycua/cua#706
• Raise clear error when using Ollama models with ComputerAgent image inputs trycua/cua#711
• Fix Windows command output decoding for UTF-8 and system encodings (#788) trycua/cua#792
• Improve Computer API Server connection warnings and timeout errors trycua/cua#803
• Windows: Detect locked desktop and return explicit input errors trycua/cua#808
• Add public computer property to RemoteDesktopSession for SDK access trycua/cua#844

Onyx (AI knowledge assistant)

• fix(mcp): pass OAuthClientProvider to call_mcp_tool for automatic token refresh onyx-dot-app/onyx#9267
• fix(llm): don't send tool_choice when no tools are provided onyx-dot-app/onyx#9224
• fix(mcp): use CE-compatible chat endpoint for search_indexed_documents onyx-dot-app/onyx#9192
• fix(open-url): preserve query params in URL lookup — fixes silent drop for URLs with query parameters onyx-dot-app/onyx#9191

All of the above are merged PRs. My work has spanned agent tooling, bug fixes, and feature improvements in these codebases. I'm comfortable navigating large TypeScript/Python repos and look forward to contributing meaningfully here.