Devika Verma 227466901e Raise the model-matrix size guardrail for the Claude 4.5/4.6 rows ...

The previous commit added Claude Opus 4.6, Opus 4.5 and Sonnet 4.5 to the matrix so
OpenWorker can be measured on those models. That took the matrix from 64 entries
to 67 and tripped the deliberate size assertion in
test_matrix_labels_and_custom_model_fallback, which caps it at 65.

Raised to 70 rather than pruning. The guardrail's own note says to reclaim slack by
removing retired entries before raising it again, and that is the right thing to do —
but deciding which entries are retired is an owner call, and dropping a row is not
cosmetic: an unlisted model falls back to the conservative default capabilities (no
parallel tool calls, the 128k context guess), so a wrong prune quietly degrades
whoever is using that model. The debt is recorded in the test comment for whoever
makes that call.

Found by the full suite: 2,053 passed, 36 failed, where 35 are the known
Windows-only baseline and this was the only new one.

2026-09-16 15:38:11 +05:30

..

corpora

OPE-136: the reviewer learns MCP names are labels, not evidence — with the eval rows to prove it

2026-08-31 08:06:14 +05:30

conftest.py

Universal scratch: every session gets a scratch root; Artifacts panel everywhere

2026-08-20 21:28:52 -07:00

test_accounts.py

OPE-136: MCP permission model — EXTERNAL floor, durable per-tool trust, run grants, honest approval plumbing

2026-08-31 08:06:14 +05:30

test_anthropic_caching.py

Enable prompt caching on the Anthropic provider

2026-07-27 22:47:30 -07:00

test_anthropic_provider.py

Fix CI: teach provider fakes messages.stream, un-Windows the write-fence probe

2026-08-31 08:48:28 +05:30

test_approval_integrity.py

Step 3b: web_search -> EGRESS + the 1.9 egress cards

2026-08-13 08:41:44 -07:00

test_artifact_walk.py

fix: stop artifact walk entering OS app-data dirs; context bar off by default

2026-07-30 13:10:30 -07:00

test_ask_user_upgrades.py

ask_user upgrades: rich options, grouped questions, option previews (OPE-51)

2026-07-29 15:52:05 +05:30

test_attachments.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_auto_approve_settings.py

Carry reviewer cache tokens into the metering badge

2026-08-22 09:51:16 +05:30

test_auto_approve.py

Persist approval provenance and server-authored mode notices

2026-08-24 14:14:51 -07:00

test_automation_create.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_automation.py

Merge pull request #50 from asklokesh/fix/automation-schedule-dst-and-weekday

2026-08-23 21:45:04 -07:00

test_autotitle.py

First contact asks via ask_user; titles read the first reply

2026-08-24 15:44:43 -07:00

test_bedrock_provider.py

Bedrock verify: detect ClientError by response shape, not class name

2026-08-17 18:57:20 +05:30

test_browser_root_scoping.py

OPE-122: enforce granted roots in browser upload and screenshot

2026-08-19 19:00:50 +05:30

test_builtin_personas.py

Toolset dedup: one windowed reader, sleep_until only, browser trim/rename

2026-08-20 23:06:36 -07:00

test_catalog.py

Toolset dedup: one windowed reader, sleep_until only, browser trim/rename

2026-08-20 23:06:36 -07:00

test_cloud_server.py

security: complete local access protections

2026-07-24 18:44:39 -07:00

test_cloud.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_code_tools.py

Keep ripgrep searches out of generated directories (#10)

2026-07-23 12:29:19 -07:00

test_codex_provider.py

Fix session auto-titling on plan-backed sessions

2026-08-24 14:48:21 -07:00

test_command_matching.py

PR3: split compound commands and check each part independently

2026-08-11 11:43:57 -07:00

test_compaction_engine.py

Make the compaction summary reliable, and stop losing the task statement (OPE-189)

2026-09-16 15:31:06 +05:30

test_compaction_persisted_summary.py

Persist the compaction record on the compacted notice and event (OPE-170 p2)

2026-09-08 21:24:36 +05:30

test_compaction_smoke.py

compaction: harden the smoke against per-turn event loops (OPE-27)

2026-07-29 18:00:38 +05:30

test_compaction_transcript.py

Make the compaction summary reliable, and stop losing the task statement (OPE-189)

2026-09-16 15:31:06 +05:30

test_compaction.py

Make the compaction summary reliable, and stop losing the task statement (OPE-189)

2026-09-16 15:31:06 +05:30

test_config.py

security: refine workspace trust controls

2026-07-24 18:44:39 -07:00

test_connections.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_connector_registry.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_connectors_allowlist.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_connectors.py

Merge branch 'main' of https://github.com/andrewyng/openworker into feature/permission-modes

2026-08-22 08:11:32 +05:30

test_conversation_atomicity.py

Make the conversation-log shrink rewrite atomic

2026-07-24 14:28:16 +09:00

test_conversation_jsonl_robustness.py

Tolerate a corrupt line when loading a conversation .jsonl

2026-07-23 14:13:07 -07:00

test_conversation_store.py

Reject path-traversal session ids in the conversation store

2026-07-23 14:09:28 -07:00

test_conversations_repair.py

fix: don't inject placeholder for trailing pending tool calls

2026-08-26 21:55:48 +05:30

test_decision_matrix_golden.py

Phase 0: golden decision table for the permission engine

2026-08-11 11:18:21 -07:00

test_devops_team.py

Shipping lineup: remove Chat, ship Code disabled, gate unshipped personas

2026-08-20 17:59:48 -07:00

test_devsecops_team.py

Shipping lineup: remove Chat, ship Code disabled, gate unshipped personas

2026-08-20 17:59:48 -07:00

test_dm_routing.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_durable_resume.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_edit_tool_guidance.py

Bound tool results, steer edits, keep compacted turns readable (OPE-186)

2026-09-15 15:09:47 +05:30

test_egress_and_overrides.py

OPE-136: MCP permission model — EXTERNAL floor, durable per-tool trust, run grants, honest approval plumbing

2026-08-31 08:06:14 +05:30

test_email_tools.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_engine_stop.py

Keep Retry offered across model switches

2026-07-23 07:43:24 -07:00

test_engine.py

Recover truncated tool calls, and never pass a leaked one off as an answer

2026-07-27 05:50:54 +05:30

test_environment.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_fake_slack.py

security: harden Slack approval handling

2026-07-24 22:33:13 -07:00

test_finish_reason.py

Persist finish_reason on assistant messages (OPE-173)

2026-09-08 16:32:11 +05:30

test_gateway_inbox_reply.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_gcal_accounts.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_gemini_provider.py

Coerce union-typed JSON Schema for Gemini tool declarations

2026-07-23 07:43:24 -07:00

test_github_installs.py

security: complete local access protections

2026-07-24 18:44:39 -07:00

test_gmail_accounts.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_hubspot_portals.py

security: complete local access protections

2026-07-24 18:44:39 -07:00

test_inbox_routing.py

Parse inbox reply intent from the leading word, not substrings

2026-07-31 14:10:37 -07:00

test_inbox.py

OPE-136: MCP permission model — EXTERNAL floor, durable per-tool trust, run grants, honest approval plumbing

2026-08-31 08:06:14 +05:30

test_interactions.py

security: harden Slack approval handling

2026-07-24 22:33:13 -07:00

test_layered_corpora.py

Add layered security corpora (gate/reviewer/sequence, 301 rows)

2026-08-19 02:35:29 +05:30

test_mangled_tool_calls.py

Diagnose truncated tool calls instead of executing their mangled args

2026-08-15 06:55:10 -07:00

test_matrix_current_models.py

Add Kimi K3 via OpenRouter to the model matrix (1,048,576 context)

2026-09-09 11:27:24 +05:30

test_max_output_tokens.py

Make the per-reply output-token limit configurable (OPE-177)

2026-09-08 17:08:29 +05:30

test_mcp_connectors.py

port aisuite#380: slack installer pre-add, mcp oauth quarantine, automation toast, e2e fixes

2026-07-21 12:56:36 -07:00

test_mcp_floor.py

Fix CI: teach provider fakes messages.stream, un-Windows the write-fence probe

2026-08-31 08:48:28 +05:30

test_mcp_oauth.py

MCP OAuth: silent refresh actually works across restarts

2026-08-21 11:44:23 -07:00

test_mcp_tool_selection.py

OPE-136: MCP permission model — EXTERNAL floor, durable per-tool trust, run grants, honest approval plumbing

2026-08-31 08:06:14 +05:30

test_mcp_trust.py

OPE-136: MCP permission model — EXTERNAL floor, durable per-tool trust, run grants, honest approval plumbing

2026-08-31 08:06:14 +05:30

test_mcp.py

MCP failure notice: one quiet line + Details + Open Connectors; stable dedupe

2026-08-21 17:18:48 -07:00

test_memory_api.py

Memory V1: remembered facts, your instructions, one screen

2026-07-28 20:32:14 +05:30

test_memory.py

Bookkeeping saves keep their place in Recents

2026-08-24 14:34:12 -07:00

test_mention_router.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_message_source.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_message_timestamps.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_model_errors.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_multiroot.py

Root promotion: request_directory gains primary; granted folder can become the workspace

2026-08-20 21:37:39 -07:00

test_openai_responses.py

Merge main into feature/permission-modes

2026-08-23 01:01:57 -07:00

test_pdf_support.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_permissions_risk.py

Step 2: Auto-Approve mode - the reviewer, the hook, and the renames

2026-08-12 12:42:17 -07:00

test_persona_connections.py

Toolset dedup: one windowed reader, sleep_until only, browser trim/rename

2026-08-20 23:06:36 -07:00

test_persona_loading.py

Personas: requires_folder/subagents/scheduling traits replace family and the workspace enum

2026-08-20 21:19:56 -07:00

test_persona_manifest.py

Personas: requires_folder/subagents/scheduling traits replace family and the workspace enum

2026-08-20 21:19:56 -07:00

test_persona_registry.py

Triage Lead: interview-first standing watch over incoming channels

2026-08-23 01:24:58 -07:00

test_persona_skills.py

personas: wire manifest skills + mcp into sessions (OPE-58)

2026-08-10 22:16:38 -07:00

test_plan_mode.py

Step 2: Auto-Approve mode - the reviewer, the hook, and the renames

2026-08-12 12:42:17 -07:00

test_project_bindings.py

Project identity: binding > git > path; named memories and boards

2026-08-23 01:24:58 -07:00

test_projects.py

Project identity: binding > git > path; named memories and boards

2026-08-23 01:24:58 -07:00

test_provenance.py

OPE-114: flag files the agent created or downloaded before running them

2026-08-19 18:20:47 +05:30

test_provider_router.py

Merge pull request #219 from hacksics/unparsed-tool-calls-upstream

2026-08-28 22:37:03 -07:00

test_provider_verify.py

feat: verify Ark credentials via Responses

2026-08-14 16:56:16 +08:00

test_providers.py

Raise the model-matrix size guardrail for the Claude 4.5/4.6 rows

2026-09-16 15:38:11 +05:30

test_readonly_grant.py

approvals: session read-only command grant; enable on consent card

2026-08-11 15:07:46 -07:00

test_readonly_scoping.py

OPE-130: scope read-only session grants to the session's folders

2026-08-22 09:16:01 +05:30

test_reasoning_effort.py

Add a reasoning_effort setting translated per provider and recorded per reply (OPE-176)

2026-09-08 20:22:10 +05:30

test_reasoning_replay.py

Replay reasoning on the OpenAI-compatible path (OPE-178)

2026-09-09 19:19:55 +05:30

test_risk_overrides.py

OPE-136: MCP permission model — EXTERNAL floor, durable per-tool trust, run grants, honest approval plumbing

2026-08-31 08:06:14 +05:30

test_run_grant.py

OPE-136: MCP permission model — EXTERNAL floor, durable per-tool trust, run grants, honest approval plumbing

2026-08-31 08:06:14 +05:30

test_secrets_file_mode.py

security: create secret files private, never chmod them after (#143)

2026-07-29 02:22:10 +05:30

test_secrets.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_security_bundles.py

Personas: requires_folder/subagents/scheduling traits replace family and the workspace enum

2026-08-20 21:19:56 -07:00

test_self_protection.py

Step 2: Auto-Approve mode - the reviewer, the hook, and the renames

2026-08-12 12:42:17 -07:00

test_self_wake.py

Toolset dedup: one windowed reader, sleep_until only, browser trim/rename

2026-08-20 23:06:36 -07:00

test_send_file.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_send_target_resolution.py

Merge pull request #488 from rkfshakti/fix/github-clone-pull-approval-gate

2026-08-23 13:05:52 -07:00

test_sender_attribution.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_served_by.py

Persist the upstream host a router reports on each reply (served_by)

2026-09-09 17:35:09 +05:30

test_server.py

Auto-title at turn start, not turn completion

2026-08-24 15:12:56 -07:00

test_session_events.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_session_facts.py

Step 1: known world frozen at session start; ingestion facts to audit only

2026-08-12 12:16:45 -07:00

test_session_persona.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_settings.py

settings: gate ollama models on a live local Ollama

2026-07-21 22:02:20 -07:00

test_shadow_eval.py

OPE-136: the reviewer learns MCP names are labels, not evidence — with the eval rows to prove it

2026-08-31 08:06:14 +05:30

test_sharing_v1.py

personas: sharing v1 — export/import bundles, version + consent (OPE-7)

2026-08-11 12:12:24 -07:00

test_shell.py

Merge pull request #216 from CyberDefenseOperations/contrib/fix-flaky-shell-background-test

2026-08-23 12:15:57 -07:00

test_skills_api.py

OPE-136: MCP permission model — EXTERNAL floor, durable per-tool trust, run grants, honest approval plumbing

2026-08-31 08:06:14 +05:30

test_skills_sessions.py

Shipping lineup: remove Chat, ship Code disabled, gate unshipped personas

2026-08-20 17:59:48 -07:00

test_skills_store.py

OPE-136: MCP permission model — EXTERNAL floor, durable per-tool trust, run grants, honest approval plumbing

2026-08-31 08:06:14 +05:30

test_skills.py

Personas: requires_folder/subagents/scheduling traits replace family and the workspace enum

2026-08-20 21:19:56 -07:00

test_slack_approval_owners.py

security: harden Slack approval handling

2026-07-24 22:33:13 -07:00

test_slack_directory.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_slack_relay.py

security: harden Slack approval handling

2026-07-24 22:33:13 -07:00

test_slack_status.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_slack_workspaces.py

security: harden Slack approval handling

2026-07-24 22:33:13 -07:00

test_standing_approvals.py

fix: claim the scheduler overlap guard at dispatch, not inside the spawned run

2026-07-31 14:08:06 -07:00

test_subagent.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_subscriptions.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_team_allowlist.py

security: harden Slack approval handling

2026-07-24 22:33:13 -07:00

test_team_board.py

fix(teams): enforce worker item visibility

2026-08-28 14:47:07 -07:00

test_team_journal.py

Drop the proposed state: boards hold only accepted work

2026-08-16 08:07:10 -07:00

test_team_open_surface.py

fix(teams): authorize attachment reads

2026-08-28 15:35:55 -07:00

test_team_store.py

fix(teams): enforce worker item visibility

2026-08-28 14:47:07 -07:00

test_team_wake.py

fix(teams): authorize attachment reads

2026-08-28 15:35:55 -07:00

test_temp_workspace.py

coworker picker: setup chips above composer, folder pick at send (UX-029)

2026-08-10 21:43:02 -07:00

test_todo_tool.py

todo_write: rename the items param to todos

2026-07-21 15:45:26 -07:00

test_token_usage.py

Fix CI: teach provider fakes messages.stream, un-Windows the write-fence probe

2026-08-31 08:48:28 +05:30

test_tool_request.py

request_tool: non-catalog names never raise the install card

2026-08-20 22:14:02 -07:00

test_tool_result_cap.py

Bound tool results, steer edits, keep compacted turns readable (OPE-186)

2026-09-15 15:09:47 +05:30

test_toolchain.py

Tool-request card: separate the product's facts from the coworker's ask

2026-08-14 23:50:51 -07:00

test_tools_permissions.py

Finish the Mode.AUTO -> Mode.BYPASS_APPROVALS rename in tests

2026-08-12 12:45:02 -07:00

test_triage_lead.py

Triage Lead: interview-first standing watch over incoming channels

2026-08-23 01:24:58 -07:00

test_truncation_continuation.py

Run both turns of the continuation reset test in one event loop (OPE-171)

2026-09-08 18:39:04 +05:30

test_tui.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_ui_refresh_e2e.py

test: fix flaky ui-refresh e2e by quiescing the auto-title call before mute assertions

2026-07-26 12:46:06 -06:00

test_unattended.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00

test_url_address_guard.py

Merge branch 'main' of https://github.com/andrewyng/openworker into feature/permission-modes

2026-08-22 08:11:32 +05:30

test_vertex_provider.py

Vertex: countTokens verify, global-location host, honest region help

2026-07-27 12:58:16 -07:00

test_wake_resume.py

Toolset dedup: one windowed reader, sleep_until only, browser trim/rename

2026-08-20 23:06:36 -07:00

test_web_search.py

OpenWorker: initial import

2026-07-21 11:09:41 -07:00