Commit Graph

  • d24f51a5fd fix(ce-code-review): back the executable floor with a total-line backstop (#1719) #1719 tmchow/review-hagfish-skill-agent Trevin Chow 2026-09-15 13:09:04 -07:00
  • 46728bc1a6 Revert "Merge compound-work phase 1 (PR #1712) into ce-optimize long-runs stack" #1708 cursor/ce-optimize-long-runs-54a9 Cursor Agent 2026-09-15 19:21:46 +00:00
  • bd709b06e0 fix(ce-code-review): recognize conventional test module names before the floor (#1719) Trevin Chow 2026-09-15 11:49:22 -07:00
  • 35d09884b0 fix(ce-code-review): address PR review feedback on the focused path (#1719) Trevin Chow 2026-09-15 11:39:52 -07:00
  • 3fb7308233 fix(ce-code-review): count renamed executables toward the full floor (#1719) Trevin Chow 2026-09-15 11:31:00 -07:00
  • dac7f2cfd7 Merge compound-work phase 1 (PR #1712) into ce-optimize long-runs stack Cursor Agent 2026-09-15 18:23:04 +00:00
  • 4bfe3d6cd3 Merge remote backend (PR #1709) into ce-optimize long-runs stack Cursor Agent 2026-09-15 18:22:55 +00:00
  • 8856cb8f57 Merge eval-discipline (PR #1707) into ce-optimize long-runs stack Cursor Agent 2026-09-15 18:22:48 +00:00
  • bedab81ce1 fix(ce-code-review): address PR review feedback (#1719) Trevin Chow 2026-09-15 11:15:49 -07:00
  • b3bdca5e0d docs(solutions): record that review cost is in entering the spine, not the findings Trevin Chow 2026-09-15 11:05:13 -07:00
  • ac4345f372 fix(ce-code-review): size review depth by consequence, not total line count Trevin Chow 2026-09-15 11:05:13 -07:00
  • 0ec66db4e9 chore: release main (#1718) compound-engineering-v3.26.2 github-actions[bot] 2026-09-14 23:57:42 -07:00
  • 2df26d7bf9 fix(ce-plan): run a Bake-off unasked when a costly choice stays open (#1717) Trevin Chow 2026-09-14 23:50:49 -07:00
  • 69fb2282e3 Address PR review feedback (#1717) #1717 tmchow/investigate-ce-bakeoff-usage Trevin Chow 2026-09-14 23:38:14 -07:00
  • dc5899648f Address PR review feedback (#1717) Trevin Chow 2026-09-14 23:26:03 -07:00
  • 5b679fa79a fix(ce-plan): run a Bake-off on its own when a costly technical choice stays open Trevin Chow 2026-09-14 23:07:03 -07:00
  • 1953002d74 chore: release main (#1716) compound-engineering-v3.26.1 github-actions[bot] 2026-09-14 21:01:42 -07:00
  • 23f6564059 fix(lfg): keep the run going after an inline child skill returns (#1715) Trevin Chow 2026-09-14 20:59:04 -07:00
  • cca97ae2fc fix(lfg): keep the run going after an inline child skill returns #1715 tmchow/fix-lfg-ce-debug Trevin Chow 2026-09-14 20:36:07 -07:00
  • 0bf4d89152 chore: release main (#1701) compound-engineering-v3.26.0 github-actions[bot] 2026-09-14 19:17:16 -07:00
  • 59a3195bba ci(windows-native): cap the job at 10 minutes (#1714) Trevin Chow 2026-09-14 17:46:00 -07:00
  • c34bf944e2 ci(windows-native): cap the job at 10 minutes #1714 ci/cap-windows-native-timeout Trevin Chow 2026-09-14 17:09:46 -07:00
  • e436e1185e docs(plans): plan the windows-native CI job cap Trevin Chow 2026-09-14 17:08:14 -07:00
  • 5a98466eb7 fix(packs): use portable guide links (#1705) Petteri Ponsimaa 2026-09-15 03:05:55 +03:00
  • f913a054bb fix(ce-code-review): align the anchor-50 escape with the code, cover two missed defect classes (#1711) Will Mitchell 2026-09-14 20:03:04 -04:00
  • 5985b5a228 chore(ce-optimize): retrigger ci Cursor Agent 2026-09-14 23:22:15 +00:00
  • 062de4b5d8 fix(ce-code-review): cheap review for small low-consequence diffs (#1706) Trevin Chow 2026-09-14 16:16:12 -07:00
  • b1322bba1d chore(ce-optimize): retrigger ci #1707 cursor/feat/ce-optimize-eval-discipline-01d6 Cursor Agent 2026-09-14 23:08:28 +00:00
  • eb7b59a2c2 chore(ce-optimize): retrigger ci Cursor Agent 2026-09-14 23:08:27 +00:00
  • 672ffa9d1f fix(test): a signalled child is a lost exit only with empty stdio (#1706) #1706 tmchow/ce-code-review-lite-depth-75af Trevin Chow 2026-09-14 16:00:29 -07:00
  • 488538858d fix(ce-code-review): protected-artifact rule binds every depth path (#1706) Trevin Chow 2026-09-14 15:52:20 -07:00
  • 5529f573c9 fix(ce-code-review): lite renders every finding and routes plan omissions (#1706) Trevin Chow 2026-09-14 15:43:19 -07:00
  • 8e7a678b76 test(knowledge): cover layout templates, knowledge-layout script, dream state, and rule parity #1712 cursor/feat-compound-work-phase1-a41d Cursor Agent 2026-09-14 22:28:47 +00:00
  • 0ba6311cc3 docs(guides): document ce-capture, ce-dream, ce-experiment and the knowledge layout Cursor Agent 2026-09-14 22:28:27 +00:00
  • 514599783a fix(ce-setup): keep ce-setup and ce-compound SKILL.md under the Codex prompt budget; accept block scalars in custom layout templates Cursor Agent 2026-09-14 22:28:27 +00:00
  • cd9442eecd feat(ce-compound): route knowledge learnings to the layout's learnings role Cursor Agent 2026-09-14 22:18:03 +00:00
  • d66dc9219b feat(ce-experiment): add thin experiment skill delegating measurement to ce-optimize Cursor Agent 2026-09-14 22:18:03 +00:00
  • 6c8d104eeb feat(ce-dream): add recurring consolidation skill with lease and fingerprint state Cursor Agent 2026-09-14 22:18:03 +00:00
  • c5b4ac9249 feat(ce-capture): add knowledge capture skill with source_id dedup and private-source boundary Cursor Agent 2026-09-14 22:18:03 +00:00
  • 813cd78d4e fix(test): treat CI spawn timeout shapes as TimeoutError Cursor Agent 2026-09-14 22:11:32 +00:00
  • b8a470d288 feat(config): add knowledge layout block, six layout templates, and ce-setup knowledge flow Cursor Agent 2026-09-14 22:09:52 +00:00
  • 95022220c9 fix(ce-code-review): give the lite receipt its finding and artifact contracts (#1706) Trevin Chow 2026-09-14 15:00:18 -07:00
  • b83a7fe2b1 fix(test): surface harness spawn timeouts as TimeoutError Cursor Agent 2026-09-14 21:57:57 +00:00
  • 58f29a1957 fix(ce-code-review): restate the depth gate as its condition; uncounted files floor (#1706) Trevin Chow 2026-09-14 14:41:21 -07:00
  • bbd9722d55 fix(ce-code-review): check repo-owned criteria on the lite path (#1706) Trevin Chow 2026-09-14 13:17:25 -07:00
  • 9516bf1fd8 fix(ce-code-review): give the lite path its receipt, plan check, and apply floor (#1706) Trevin Chow 2026-09-14 11:52:14 -07:00
  • 9f0aa9c7fe docs(ce-optimize): document the remote backend and cross-machine comparability #1709 cursor/ce-optimize-long-runs-remote-54a9 Cursor Agent 2026-09-14 18:08:48 +00:00
  • 9bb93861ea feat(ce-optimize): remote experiment backend with paired self-measurement Cursor Agent 2026-09-14 18:08:48 +00:00
  • 6d7057f308 docs(ce-optimize): name the wake condition for the required holdout Cursor Agent 2026-09-14 18:13:31 +00:00
  • ea9d671b46 docs(ce-optimize): harness validity, holdout, and instruction-text guidance Cursor Agent 2026-09-14 17:50:53 +00:00
  • 8c0725b8d6 feat(ce-optimize): held-out confirmation, judge validity gate, and text-target guidance Cursor Agent 2026-09-14 17:50:53 +00:00
  • 72aee4b016 docs(ce-optimize): document long runs, the state root, and the wrap-up report Cursor Agent 2026-09-14 18:00:11 +00:00
  • 217a516194 feat(ce-optimize): resumable tick loop with durable state root Cursor Agent 2026-09-14 18:00:11 +00:00
  • eb46ea65dc fix(ce-code-review): size lite from helper facts plus consequence Cursor Agent 2026-09-14 17:47:43 +00:00
  • fd8abda7f6 feat(lfg): route each request to the skill that owns it before shipping (#1702) Trevin Chow 2026-09-13 19:41:45 -07:00
  • a2a8643800 fix(lfg): defect route needs a fix request; attendance is a person who can answer #1702 tmchow/improve-lfg-skill-routing Trevin Chow 2026-09-13 18:56:27 -07:00
  • e95f07517e fix(lfg): defect route takes concrete reports; brainstorm grounding reaches ce-plan Trevin Chow 2026-09-13 18:48:36 -07:00
  • dc48c477b4 fix(lfg): round-5 review restatements on shipping, brainstorm, and prototype handoffs Trevin Chow 2026-09-13 18:41:08 -07:00
  • f58f4677d8 fix(lfg): round-4 review restatements on the routed front half Trevin Chow 2026-09-13 18:32:37 -07:00
  • f7a5e43097 fix(lfg): state the offered-work block as a goal; ce-debug returns facts, not verdicts Trevin Chow 2026-09-13 18:23:12 -07:00
  • a54024af26 fix(lfg): a plan path alone does not select the plan route Trevin Chow 2026-09-13 18:10:04 -07:00
  • 35358b19ca fix(lfg): a judgment-only request ends at the verdict; offered work means an open PR contains it Trevin Chow 2026-09-13 17:58:03 -07:00
  • 3435b59148 fix(lfg): ship only what the user offered on the defect route; let an attended run reach ce-prototype Trevin Chow 2026-09-13 17:49:07 -07:00
  • 180b1186c8 docs(lfg): describe the routed front half and record the proposal and eval trail Trevin Chow 2026-09-13 17:31:40 -07:00
  • 76c4e5841c feat(lfg): route each request to the skill that owns it before shipping Trevin Chow 2026-09-13 17:31:40 -07:00
  • aae9f91c01 fix(ce-noslop): keep the edit-mode change summary out of user-facing text (#1700) Trevin Chow 2026-09-13 16:44:35 -07:00
  • 59c2304bf8 docs(skill-design): record the caller-only channel rule in the authoring standard #1700 tmchow/fix-ce-noslop-prose Trevin Chow 2026-09-13 16:15:36 -07:00
  • a08bddba09 fix(ce-noslop): let a requested change line reach its requester Trevin Chow 2026-09-13 16:10:25 -07:00
  • dc16d29782 fix(ce-noslop): keep the edit-mode change summary out of user-facing text Trevin Chow 2026-09-13 15:59:10 -07:00
  • 44d65ad64a chore: release main (#1610) compound-engineering-v3.25.0 github-actions[bot] 2026-09-12 20:44:58 -07:00
  • 53af1a2eab fix(ce-code-review): promote agreement only with a verified cross-model peer (#1697) Trevin Chow 2026-09-12 17:27:35 -07:00
  • 77f8824dd2 fix(ce-code-review): carry the unmeasured-incidence caveat in validation_reason (#1697) #1697 tmchow/review-same-model-promotion Trevin Chow 2026-09-12 17:03:40 -07:00
  • f68a476fcb fix(ce-code-review): promote agreement only with a verified cross-model peer Trevin Chow 2026-09-12 16:56:58 -07:00
  • 98ca50b719 fix(ce-code-review): count .mjs and .cjs files as executable changed lines (#1696) Trevin Chow 2026-09-12 16:16:52 -07:00
  • e84a4b08c5 fix(ce-code-review): count .mjs and .cjs files as executable changed lines #1696 tmchow/ce-debug-issue-1666 Trevin Chow 2026-09-12 16:12:27 -07:00
  • ea9d090b54 feat(ce-code-review): require cited evidence to reject a protected-subject finding (#1694) Chris McCarty 2026-09-12 16:04:47 -07:00
  • 3f94834661 fix(ce-code-review): apply the protected-subject veto in the report leaf #1695 tmchow/ce-code-review-protected-subject-veto-47ca Cursor Agent 2026-09-12 21:20:57 +00:00
  • 35eb71e660 fix(ce-code-review): block a merge-ready verdict while an unresolved P0 or P1 gate remains Chris McCarty 2026-09-12 12:06:18 -07:00
  • e2f9cd3d09 feat(ce-code-review): require cited evidence to reject a protected-subject finding Chris McCarty 2026-09-12 11:00:00 -07:00
  • 7511114eec fix(ce-code-review): finish the round from run-dir artifacts in a fresh context (#1692) Trevin Chow 2026-09-12 13:35:41 -07:00
  • 872b53ef46 fix(ce-code-review): delegate remote inspection to diff-scope.md; persist in-band verdicts #1692 tmchow/issue-1690 Trevin Chow 2026-09-12 13:15:15 -07:00
  • ea8758341a fix(ce-code-review): state the handoff condition; carry constraints, plan extraction, remote-scope rule Trevin Chow 2026-09-12 13:06:28 -07:00
  • bbbca1c75d fix(ce-code-review): materialize handoff paths, keep base: apply-eligible, load conventions before apply Trevin Chow 2026-09-12 12:57:56 -07:00
  • eaf8ca3db0 fix(ce-code-review): record every validator outcome for the report leaf Trevin Chow 2026-09-12 12:49:44 -07:00
  • 2129a43ab4 fix(ce-code-review): the dispatch context owns the peer through fold-in; leaves only consume it Trevin Chow 2026-09-12 12:44:20 -07:00
  • 947a8bda2a fix(ce-code-review): make mode.apply_local the only apply authority inside a leaf Trevin Chow 2026-09-12 12:33:11 -07:00
  • 18b3cf2d82 fix(ce-code-review): finish through two leaf delegates; the orchestrator launches every subagent Trevin Chow 2026-09-12 12:30:52 -07:00
  • 5dee03b541 fix(ce-code-review): inline the finish handoff in the finish subagent's prompt Trevin Chow 2026-09-12 11:25:36 -07:00
  • 580b8e7bca fix(ce-code-review): keep peer-recipient changes out of the finish context Trevin Chow 2026-09-12 11:12:52 -07:00
  • 6b4ef1f682 fix(ce-code-review): finish the round from run-dir artifacts in a fresh context Trevin Chow 2026-09-12 11:05:17 -07:00
  • 8bc0d03f55 fix(ce-code-review): bound the Stage 4 reviewer wait on each reviewer's artifact (#1691) Trevin Chow 2026-09-12 10:46:04 -07:00
  • 961a4b276e fix(ce-code-review): stop a reviewer whose artifact landed while its launch is live #1691 tmchow/issue-1689 Trevin Chow 2026-09-12 10:28:03 -07:00
  • 1d699b8eb1 fix(ce-code-review): launch reviewers with a bounded collector; clock each from its own launch Trevin Chow 2026-09-12 10:21:03 -07:00
  • 22d40f04c8 fix(ce-code-review): bound the Stage 4 reviewer wait on each reviewer's artifact Trevin Chow 2026-09-12 10:12:43 -07:00
  • 25b9c90e5d fix(ce-code-review): bound the validator wait on an on-disk verdicts file (#1688) Trevin Chow 2026-09-12 09:11:00 -07:00
  • 9c1bdeda22 ci: cap the test job at 30 minutes (#1687) Trevin Chow 2026-09-12 09:10:45 -07:00
  • a6149080e8 fix(ce-code-review): reach the validator bound with repeated waits #1688 tmchow/debug-issue-1679 Trevin Chow 2026-09-12 08:47:29 -07:00
  • fe9af3261a fix(ce-code-review): bound the validator wait on an on-disk verdicts file Trevin Chow 2026-09-12 08:41:42 -07:00
  • a8bfc3954b ci: cap the test job at 30 minutes #1687 tmchow/ci-test-timeout Trevin Chow 2026-09-12 08:39:55 -07:00
  • d484685805 fix(skill-eval): grade declared decisions instead of one host's wording (#1686) Trevin Chow 2026-09-12 08:39:17 -07:00