dongarandClaude Opus 5.5 8bd1ac0fa2 Adopt an already-published PR instead of publishing again ...

Fixes counterexample 5 from specs/orchestrator.qnt. A crash after the
PR was opened but before PR_OPEN was recorded left the agent in
EXECUTING; the restart question's 'continue' and 'start' then published
again, Gitea refused a second PR, and there was no way back to PR_OPEN.
On the live run the executor model happened to notice the existing
work; recovery must not depend on that.

- Boot (A3 step 4): in EXECUTING, if Gitea shows the milestone's PR
  open, adopt it, watch it from the start, and resume in PR_OPEN with a
  notice. Gitea unreachable or the PR closed: the restart question, as
  before.
- MilestonePublisher.publish is idempotent: after pushing, an open PR
  for the branch is adopted (and so updated) rather than a second one
  opened. This also covers a PR opened whose response was lost.
- The model adopts too, gains the lost-response outcome, and now counts
  PRs opened per milestone (one_pr_per_milestone was a flag the fixed
  model could never set). A mutant without adoption fails the new
  scenario; the properties held over 200,000 traces each.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE

2026-09-29 19:58:19 +00:00

..

test_abandon.py

Add the abandon command

2026-09-29 17:43:43 +00:00

test_agent_build.py

Enforce the tool-call contract, asking servers only where it works

2026-09-24 22:54:49 +00:00

test_approval_binding.py

Bind approvals and feedback to the revision they were about

2026-09-29 19:13:20 +00:00

test_approvals.py

Externalise plans to Gitea and accept approvals there (A4, Part H)

2026-09-07 22:15:08 +00:00

test_bench.py

Add four harder benchmark tasks

2026-09-27 22:43:39 +00:00

test_canonical.py

Add ask_expert and eval capture (A14/A16) — step 7c

2026-09-08 19:48:13 +00:00

test_catalog.py

Discover a model's vendor from LiteLLM instead of its name

2026-09-23 01:47:58 +00:00

test_config.py

Use qwen3.8-27b as the example executor

2026-09-29 14:18:50 +00:00

test_context.py

Enforce the tool-call contract, asking servers only where it works

2026-09-24 22:54:49 +00:00

test_control.py

Bind approvals and feedback to the revision they were about

2026-09-29 19:13:20 +00:00

test_delegation.py

Add the operator kill switch, stop (A4)

2026-09-23 22:47:29 +00:00

test_dev_main.py

Make the config file the only home for settings

2026-09-23 23:25:58 +00:00

test_evals.py

Add ask_expert and eval capture (A14/A16) — step 7c

2026-09-08 19:48:13 +00:00

test_executor.py

Enforce the tool-call contract, asking servers only where it works

2026-09-24 22:54:49 +00:00

test_expert.py

Add ask_expert and eval capture (A14/A16) — step 7c

2026-09-08 19:48:13 +00:00

test_gates.py

Bind approvals and feedback to the revision they were about

2026-09-29 19:13:20 +00:00

test_git.py

Publish work the executor committed itself, and gate when there is none

2026-09-29 16:17:41 +00:00

test_gitea.py

Add the abandon command

2026-09-29 17:43:43 +00:00

test_llm.py

Record rate-limit waits in benchmark runs, and leave them out of time

2026-09-27 20:57:15 +00:00

test_matrix.py

Handle leading mentions in Matrix messages

2026-09-29 15:59:27 +00:00

test_no_stranding.py

Never leave the agent stranded in a working state

2026-09-29 16:39:20 +00:00

test_oplog.py

Add literal-format test showing exact log lines

2026-09-05 21:20:41 +00:00

test_planner.py

Enforce the token cap, and treat a spent budget as an interruption

2026-09-22 23:54:57 +00:00

test_poller.py

Add PR comment intake and the review loop (B2) — step 6b

2026-09-07 21:59:50 +00:00

test_pr_open_handling.py

Keep Gitea threads watched across restarts and interruptions

2026-09-29 17:33:41 +00:00

test_pr_open.py

Publish work the executor committed itself, and gate when there is none

2026-09-29 16:17:41 +00:00

test_probe.py

Enforce the tool-call contract, asking servers only where it works

2026-09-24 22:54:49 +00:00

test_publish.py

Adopt an already-published PR instead of publishing again

2026-09-29 19:58:19 +00:00

test_quint_counterexamples.py

Adopt an already-published PR instead of publishing again

2026-09-29 19:58:19 +00:00

test_reconcile.py

Implement A9 state-change logging in the state store

2026-09-05 21:13:36 +00:00

test_render_anthropic.py

Add the Anthropic passthrough renderer and client (A10/A13) — 7a

2026-09-07 23:13:22 +00:00

test_render.py

Render one assistant message per turn, and make the request timeout a setting

2026-09-27 22:15:30 +00:00

test_review_loop.py

Add PR comment intake and the review loop (B2) — step 6b

2026-09-07 21:59:50 +00:00

test_state.py

Implement A9 state-change logging in the state store

2026-09-05 21:13:36 +00:00

test_stop.py

Bind approvals and feedback to the revision they were about

2026-09-29 19:13:20 +00:00

test_tools.py

Fit every model request to its context window

2026-09-24 00:26:30 +00:00

test_worker.py

Add the operator kill switch, stop (A4)

2026-09-23 22:47:29 +00:00