browser-use · Issues· 435 open

Bug: Depreciated syntax used to list development dependencies in pyproject.toml

Bug: schema cleanup removes fields named default, minItems, or min_items

allowed_domains / prohibited_domains only gate agent-initiated navigation; every other route to a page skips the gate

Bug: user-agent shadow roots are reported as SHADOW(open) in the serialized page state

Domain control plane: scheme-qualified allowed_domains entries are prefix-matched; default profile grants clipboardReadWrite to every origin

Skill output schemas ignore enum constraints

js() runner silently returns empty results for entire session (Runtime.evaluate detach?) — undetectable dead state

Bug: GIF overlay crashes with display_step=False (UnboundLocalError: y_step)

CLI 3.0 harness: Input.dispatchMouseEvent/key events silently dropped in a long-lived session while Runtime.evaluate stays healthy (stale Input session; ensure_real_tab doesn't recover)

select_dropdown does not handle readonly inputs backed by clickable native listboxes

Preserve filled state for redacted password inputs in DOM snapshots

Reuse history objects before querying the LLM, with LLM fallbackReuse existing history objects whenever they contain the information needed for the current task, instead of querying the LLM again. The history should be checked first and used when a suitable object is available. If no relevant history object exists, or the history does not contain sufficient information, fall back to querying the LLM as it does today. This would reduce redundant LLM calls while preserving the existing behavior as a fallback.

Bug: evaluate hides JavaScript exception details behind "Uncaught"

Feature Request: RFC: Standardized Evaluation/Benchmarking Harness for LLM Tool Calling

Bug: ...Performance degradation during long-running agent tasks due to repeated DOM serialization, browser state fetching, and growing message history