L/docs updates agent

[lorenss-m]

Note

Medium Risk
Medium risk because hud sync tasks diff/signature logic and upload payloads now include custom columns and inferred column schemas, which can change what gets uploaded and how the platform interprets task metadata; the rest is largely documentation reorganization.

Overview
CLI: hud sync tasks now supports per-task custom columns that participate in diff signatures, upload as column_values, and auto-infer/merge evalset column schemas; it also adds --export to write remote tasks to .json or flattened .csv.

SDK/agents: Adds Task.columns to the task model and improves missing-API-key error messages in ClaudeAgent, OpenAIAgent, and GeminiAgent.

Docs: Major reorg—removes older “concepts/testing/best practices/deploy/environments/tools index” pages, adds new “Building” pages (scaffolding, tasks-and-evaluation, running-at-scale, environments-as-data), expands advanced/patterns to cover sandboxing/mocking, and updates site navigation and cross-links accordingly.

^{Reviewed by Cursor Bugbot for commit 573ab64. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 5dc58fa. Configure here.}

[cursor]

Unreachable single-select branch due to inference gap

Low Severity

_infer_column_type documents returning "single-select" but only ever returns "text", "number", or "multi-select". This makes the if col_type == "single-select" branch in _build_column_definitions unreachable dead code. String columns that could be single-select (e.g., ["easy", "medium", "hard"]) will always be inferred as "text" and won't receive options metadata.

Additional Locations (1)

• hud/cli/sync.py#L301-L303

 

^{Reviewed by Cursor Bugbot for commit 5dc58fa. Configure here.}