81 Commits

Author SHA1 Message Date
John Preston
aec4539f90 [ai] Match the completion marker as a whole log line
Task: 2026/09/02/match-the-test-complete-marker-as-a-whole-line
2026-09-03 22:26:26 +04:00
John Preston
97bc3dd39e [ai] Record MTP retries, space-class reads and N/A rows
Task: 2026/09/02/add-code-500-probe-space-class-check-and-na-rows-to-the-harness
2026-09-03 22:26:26 +04:00
John Preston
f4c4e4d322 [ai] Don't read all tasks on consolidation phase. 2026-09-01 11:21:14 +04:00
John Preston
735c50f351 [ai] Stop routing platform checks for build breakage
Every platform is built before anything merges, so a build or link break is
found there instantly. A platform follow-up now needs platform-specific code
and a concrete reason to suspect that platform.
2026-08-28 10:41:17 +04:00
John Preston
d8708e6be1 [ai] Do not review a task that wrote no code
The initial five-lens fanout applied unconditionally, so an already-satisfied verification task convened six reviewers with no diff to read and they reviewed its measurement contract instead. Restrict the fanout to tasks that produced a source diff, bound the no-diff path to one review round, and send contract defects to the evidence loop as check repairs.
2026-08-28 10:40:46 +04:00
John Preston
27b0947e88 [ai] Route follow-ups beside the active task 2026-08-28 10:40:46 +04:00
John Preston
c513f8c8ed [ai] Verify cross-platform code once
Treat another platform as an Unverified: gap only when the diff carries a platform-sensitive mechanism, and drop a second-host re-run of portable behaviour in the discovered-follow-up scope filter.
2026-08-28 10:40:46 +04:00
John Preston
16360c603a [ai] Refuse implausible test watchdog values
Task: 2026/08/23/refuse-implausible-test-watchdog-values
2026-08-28 10:40:46 +04:00
John Preston
8e7bdb9967 [ai] Keep debug machinery out of production units 2026-08-24 11:02:55 +04:00
John Preston
a94bc4168d [ai] Debug startup crashes before blaming fixtures 2026-08-24 11:02:47 +04:00
John Preston
d428defec0 [ai] Route oversized tasks through durable splits 2026-08-24 11:02:10 +04:00
John Preston
03f59cd642 [ai] Fan out initial task reviews across all lenses 2026-08-24 11:02:10 +04:00
John Preston
a49c572f77 [ai] Keep task review and testing focused and convergent 2026-08-24 11:02:10 +04:00
John Preston
1fd1682934 [ai] Retire review lenses from the diff, not the plan 2026-08-21 21:23:16 +04:00
John Preston
ee3d9d1aa3 [ai] Scope AI prefix to workflow-only commits 2026-08-21 17:10:31 +04:00
John Preston
f022cec701 [ai] More decisions to agents in workflow.
Task: 2026/08/20/add-tooling-evidence-profile-to-perform-task
2026-08-21 17:10:30 +04:00
John Preston
85a62495f2 [ai] Read helper paths literally, fix crash detection
Task: 2026/08/18/repair-helper-path-reads-and-linux-crash-diagnosis
2026-08-18 21:50:09 +04:00
John Preston
c367454e7b [ai] Read non-ASCII paths literally in workspace scans
Task: 2026/08/18/read-non-ascii-paths-literally-in-change-scans
2026-08-18 21:50:09 +04:00
John Preston
02e51c10ed [ai] Stage removed paths in workspace helper commits
Task: 2026/08/18/stage-removed-paths-in-helper-commits
2026-08-18 21:50:09 +04:00
John Preston
449b2eba55 [ai] Digest retained task content from git, not the checkout
A consolidation alias sealed on one platform could never revalidate on
another, because the seal hashed working-tree bytes.
2026-08-18 18:15:11 +04:00
John Preston
d80ca46165 [ai] Improve task test harness reliability and guidance. 2026-08-16 22:17:49 +04:00
John Preston
0cd2b4fcae [ai] Keep test-flaw recovery inside task runs 2026-08-16 14:02:42 +04:00
John Preston
d4319b82ff [ai] Write out the model name finishing the task. 2026-08-14 17:17:03 +04:00
John Preston
a8def18edc [ai] Gate tasks on source branch lineage 2026-08-12 22:28:05 +04:00
John Preston
48046aa140 [ai] Assert build-lock selection reasons
Task: 2026/08/11/assert-build-lock-reasons-not-just-pids
2026-08-12 22:28:05 +04:00
John Preston
ef295c8041 [ai] Test workflow in Grok Build. 2026-08-12 22:28:05 +04:00
John Preston
22ab8c35a9 [ai] Repair three macOS-only harness suite tests
Task: 2026/08/10/repair-macos-only-harness-suite-failures
2026-08-11 15:35:48 +04:00
John Preston
fdae24e725 [ai] Report a death after TEST_COMPLETE in test-run
Task: 2026/08/10/report-post-complete-death-in-test-run
2026-08-11 15:22:34 +04:00
John Preston
00c484f34a [ai] Add a minimal task type to the AI workflow 2026-08-10 18:42:47 +04:00
John Preston
1d7a6d5ff2 [ai] Consolidate discovered follow-ups. 2026-08-10 18:12:32 +04:00
John Preston
ffd27d66bc [ai] Add performance review lens. 2026-08-10 18:12:32 +04:00
John Preston
4a8d7fc7d9 Make test recovery more direct
Escalate repeated setup failures through direct production seams and preserve disposable submodule probes.
2026-08-04 13:08:47 +04:00
John Preston
e65ea19e46 Defer postponed calls during harness synthetic input
Task: 2026/07/31/make-harness-input-drive-postponed-fixups
2026-08-04 11:50:39 +03:00
John Preston
f7c00e667c Make harness synthetic input settle postponed fix-ups
Task: 2026/07/31/make-harness-input-drive-postponed-fixups
2026-08-04 11:50:39 +03:00
John Preston
4573d38b85 [ai] Remove test code from release. 2026-07-31 22:53:32 +04:00
John Preston
ccab70a8cd Make chats-list wait non-fatal
Task: 2026/07/31/survive-unloaded-chats-list-in-test-harness
2026-07-31 19:11:29 +04:00
John Preston
69527e8be9 Recover autonomous builds from output locks 2026-07-31 19:11:29 +04:00
John Preston
b23c32dd43 Reconcile harness docs with stale crash setup
Task: 2026/07/30/reconcile-harness-docs-with-setup-contract
2026-07-31 19:11:29 +04:00
John Preston
798fc5c970 [ai] Diagnose a crash after clearing stale state
Task: 2026/07/29/clear-stale-crash-report-before-test-run
2026-07-30 19:55:13 +04:00
John Preston
08f466f719 [ai] Clear a stale crash report before a test run
Task: 2026/07/29/clear-stale-crash-report-before-test-run
2026-07-30 19:55:10 +04:00
John Preston
668eccc03b [ai] Preserve project lineage in task routing. 2026-07-30 09:56:22 +04:00
John Preston
ca5ac774ec [ai] Clear stale crash dumps in test SETUP 2026-07-29 20:29:48 +04:00
John Preston
eb834033be [ai] Attempt to judge the structure of the plan. 2026-07-29 12:06:55 +04:00
John Preston
ce7790427a [ai] Skip code review when there is no product diff 2026-07-28 16:05:57 +04:00
John Preston
64c8d95ab9 [ai] Improve testing cycles. 2026-07-28 14:31:26 +04:00
John Preston
140c9a93fd [ai] Treat the live test folder as disposable 2026-07-28 12:27:33 +04:00
John Preston
87a022ad1d [ai] Have semi-prepared testing run. 2026-07-28 10:55:01 +04:00
John Preston
4815f89bd4 [ai] Keep Computer Use interruptions non-blocking. 2026-07-27 15:26:26 +04:00
John Preston
807ff98e8b [ai] Review from independent lenses and route unverified gaps 2026-07-27 15:26:26 +04:00
John Preston
d853721afe [ai] Protect portable data during Xcode clean rebuilds 2026-07-24 17:02:14 +04:00