All pull requests

Pull requests

• fix(prime): stage pinned release dependencies from GitHub

  area:agentbugneeds-review
  #6208·yaoyu-33 opened Nov 15, 2026Contributor·

  ·
  1

• fix(nooa): grade the native text-only stop

  area:agentbugneeds-review
  #6207·yaoyu-33 opened Nov 15, 2026Contributor·

  ·
  1

• fix(droid): verify work after native repetition stops

  area:agentbugneeds-review
  #6206·yaoyu-33 opened Nov 15, 2026Contributor·

  ·
  1

• fix(codex): retain native usage on budget stops

  area:agentbugneeds-review
  #6205·yaoyu-33 opened Nov 15, 2026Contributor·

  ·
  1

• fix(hermes): classify exhausted continuations as output limits

  area:agentbugneeds-review
  #6204·yaoyu-33 opened Nov 15, 2026Contributor·

  ·
  1

• fix(cli): reject ignored native harness selections

  area:clibugneeds-review
  #6203·yaoyu-33 opened Nov 15, 2026Contributor·

  ·
  1

• Wchai/finance agent resources retry

  #6202·wowowoxuan1994 opened Nov 15, 2026Contributor·7 tasks done·

  ·
  19

• feat(gdpval): add OpenHands training harness and disaggregated judge

  area:agentfeatureneeds-review
  #6201·yaoyu-33 opened Nov 15, 2026Contributor·

  ·
  15

• feat(vllm_model): add opt-in tool_parameter_types for calls to undeclared tools

  area:modelcomplexity:lowfeatureneeds-review
  #6200·linj-glitch opened Nov 15, 2026Collaborator·7 tasks done·

  ·
  1

• fix(server): pass the server config by file instead of an env var

  #6199·cmunley1 opened Nov 14, 2026Contributor·

  ·
  1

• fix(deepswe): score tests that hit their time limit as unresolved

  #6198·AdamRajfer opened Nov 14, 2026Contributor·

  ·
  3

  

• fix(deepswe): stage a pinned ripgrep into the OpenCode agent sandbox

  #6197·AdamRajfer opened Nov 14, 2026Contributor·

  ·
  1

  2

  

• fix(prefix-pass-k): mask failed commits and support model workers

  #6196·ko3n1g opened Nov 14, 2026Contributor·

  ·
  3

• fix(opencode_agent): fail sandbox setup when ripgrep is missing

  #6195·AdamRajfer opened Nov 14, 2026Contributor·

  ·
  

• test(crewai): give the deadline tests time for the crew to start on a loaded runner

  #6193·AdamRajfer opened Nov 14, 2026Contributor·

  ·
  4

  

• fix(swebench): print per-test results when pytest runs without -rA

  #6192·AdamRajfer opened Nov 14, 2026Contributor·

  ·
  6

  

• feat(collab_intel): run the policy's code in a sandbox

  #6191·cmunley1 opened Nov 14, 2026Contributor·

  ·
  10

• fix(server): skip launch entries without an entrypoint

  internal_sync
  #6190·yaoyu-33 opened Nov 14, 2026Contributor·

  ·
  1

• fix(rollouts): preserve predecessor identity in projected reads

  internal_sync
  #6189·yaoyu-33 opened Nov 14, 2026Contributor·

  ·
  1

• feat(health): flag tool calls left in the answer by a mismatched parser

  #6188·AdamRajfer opened Nov 14, 2026Contributor·

  ·
  2

  

• Make OpenCode the default for visual benchmark catalogs

  #6187·katerynaCh opened Nov 14, 2026Contributor·

  ·
  11

• refactor(resources)!: generalize training tool-reward wrappers

  #6186·katerynaCh opened Nov 14, 2026Contributor·6 of 7 tasks·

  ·
  16

• nl2repobench follow-ups: leak fixes, trace_only, OpenCode termination/idle nudge, prompt parity

  community-request
  #6184·drekesh-cell opened Nov 14, 2026·6 of 7 tasks·

  ·
  1

• fix(local_vllm_model): restore vLLM runtime deps the shared policy never-installs

  QA_fix
  #6183·qiaochuz-nv opened Nov 14, 2026Contributor·7 tasks done·

  ·
  2

• perf(gdp_pdf_agent): render documents off the event loop

  area:agent
  #6182·yaoyu-33 opened Nov 14, 2026Contributor·

  ·
  2