From 13e3db585d6eea5261df526a33c874106f84c571 Mon Sep 17 00:00:00 2001 From: Daniel McCoy Stephenson Date: Sat, 14 Nov 2026 23:35:00 -0700 Subject: [PATCH 1/3] Check the default branch's current CI state at Phase 1 Co-Authored-By: Claude Opus 5.5 --- RESEARCH.md | 1 + create-dev-loop.md | 2 ++ 2 files changed, 3 insertions(+) diff --git a/RESEARCH.md b/RESEARCH.md index f2c60af..7352bb2 100644 --- a/RESEARCH.md +++ b/RESEARCH.md @@ -52,6 +52,7 @@ Last reviewed: 2026-08-04. - PR #121 (Define the structurally-red-anchor path and require the anchor to have executed tests): shipped 2026-08-16. Observed effect: pending — needs N cycles of data. Extends the same argument as PR #54's "green CI is not verification when CI's scope excludes the changed files" to two further ways an anchor can carry no information: a required job that has *never* succeeded on the base branch (observed on a repo whose CI jobs have failed on empty credentials since February 2026, so every cycle terminated in a hand-off regardless of quality), and a build tool that reports success having executed zero tests (`:test NO-SOURCE` still prints `BUILD SUCCESSFUL`). Both were previously read as ordinary signals — the first as a blocking red, the second as a green — which is unanchored critique wearing an anchor's clothes. - PR #136 (Make the checkout form Phase 4's default Tests-fix revert and gate ladder rung 1): shipped 2026-08-27. Observed effect: pending — needs N cycles of data. Repairs the mechanism PR #54's entry above introduced. The prescribed `git stash push` revert ran against a tree Phase 3 had already committed and Phase 4 had already pushed, so it saved nothing, the check then passed against the still-fixed tree, and the no-op was naturally scored as a false-negative in the *test* — an anchor reporting a verdict it never measured, which is precisely the failure this finding warns about. The checkout form becomes the default because it is correct in the state Phase 4 actually finds, and the revert must now be confirmed to have changed the tree before either half is trusted. Separately, the ladder rung PR #67 added is gated against headless dispatch, where an interrupted session leaves the branch head undoing its own fix and the substitute anchor damages the thing it substitutes for. - PR #138 (Ground Phase 4's Docs score in a real pass): shipped 2026-08-29. Observed effect: pending — needs N cycles of data. The **Docs** rubric item named a check that Phase 7 did not perform until three phases later, so the only way to score it on time was from judgment — an anchor-shaped item with no anchor behind it, the same defect as PR #121's zero-executed-tests green, relocated from the signal to the schedule. The table pass now runs at Phase 4, where the score is taken, and Phase 7 is reduced to a re-check of whatever the review phases changed. Phase 7 is also required to record that it ran, since a clean sweep that leaves no artifact is indistinguishable from a skipped one — which is the auditability half of the same problem. +- PR #TBD (Check the default branch's current CI state and take a pre-edit baseline at Phase 1): shipped pending. Observed effect: pending — needs N cycles of data. PR #121 covered a required job that has *never* passed on the base branch, but not a job that passes routinely and is red at the moment. That case was seen in a generated instance whose `main` had failed its `Build` check for five days without any cycle noticing (#144). A PR branched off a red base inherits the red check, so the anchor says nothing about the change under review. Running the anchor once before editing means a later failure can be told apart from one the edit caused. --- diff --git a/create-dev-loop.md b/create-dev-loop.md index af24486..3ae2d50 100644 --- a/create-dev-loop.md +++ b/create-dev-loop.md @@ -185,6 +185,8 @@ gh run list --workflow --branch {{DEFAULT_BRANCH}} --limit 50 -- \`\`\` If a required workflow shows no successful run in its whole recorded history on `{{DEFAULT_BRANCH}}`, record the repo as **structurally red** for that workflow, together with the cause (absent credentials, an unset secret, an unavailable license) and any issue tracking it. This is a fact about the repository, not about the change — establishing it once at triage is what lets Phase 4 assess per-job signal and Phase 8 name the blocking condition, instead of rediscovering it at the merge gate every cycle. +**Check whether `{{DEFAULT_BRANCH}}` is red right now.** A workflow that succeeds routinely can still be failing at the moment, and the structurally-red check above does not catch that, because the workflow *has* passed before. Read the most recent run of each required workflow in the same `gh run list --branch {{DEFAULT_BRANCH}}` output. If the latest run of a required check on `{{DEFAULT_BRANCH}}` failed, that breakage is this cycle's work and outranks the open-issue backlog. While it stands, every PR inherits the red check, so the Phase 4 external anchor says nothing about the change under review. Before editing anything, run the Phase 3 verify commands once on an unmodified `{{DEFAULT_BRANCH}}` as a baseline, even when `{{DEFAULT_BRANCH}}` is green. A failure first seen after editing cannot be told apart from one the edit caused, and the cycle will spend turns proving it did not. + **Close stale-open issues first.** Check whether any open issues were resolved by recent PRs but not yet closed. Cross-reference `git log` against open issue titles: \`\`\`bash From 3c643a20f1f20b80d0fc45875b51917891a7c46f Mon Sep 17 00:00:00 2001 From: Daniel McCoy Stephenson Date: Sat, 14 Nov 2026 23:36:13 -0700 Subject: [PATCH 2/3] Say what to do when the default branch's breakage is out of reach Co-Authored-By: Claude Opus 5.5 --- create-dev-loop.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/create-dev-loop.md b/create-dev-loop.md index 3ae2d50..3da50b4 100644 --- a/create-dev-loop.md +++ b/create-dev-loop.md @@ -185,7 +185,7 @@ gh run list --workflow --branch {{DEFAULT_BRANCH}} --limit 50 -- \`\`\` If a required workflow shows no successful run in its whole recorded history on `{{DEFAULT_BRANCH}}`, record the repo as **structurally red** for that workflow, together with the cause (absent credentials, an unset secret, an unavailable license) and any issue tracking it. This is a fact about the repository, not about the change — establishing it once at triage is what lets Phase 4 assess per-job signal and Phase 8 name the blocking condition, instead of rediscovering it at the merge gate every cycle. -**Check whether `{{DEFAULT_BRANCH}}` is red right now.** A workflow that succeeds routinely can still be failing at the moment, and the structurally-red check above does not catch that, because the workflow *has* passed before. Read the most recent run of each required workflow in the same `gh run list --branch {{DEFAULT_BRANCH}}` output. If the latest run of a required check on `{{DEFAULT_BRANCH}}` failed, that breakage is this cycle's work and outranks the open-issue backlog. While it stands, every PR inherits the red check, so the Phase 4 external anchor says nothing about the change under review. Before editing anything, run the Phase 3 verify commands once on an unmodified `{{DEFAULT_BRANCH}}` as a baseline, even when `{{DEFAULT_BRANCH}}` is green. A failure first seen after editing cannot be told apart from one the edit caused, and the cycle will spend turns proving it did not. +**Check whether `{{DEFAULT_BRANCH}}` is red right now.** A workflow that succeeds routinely can still be failing at the moment, and the structurally-red check above does not catch that, because the workflow *has* passed before. Read the most recent run of each required workflow in the same `gh run list --branch {{DEFAULT_BRANCH}}` output. If the latest run of a required check on `{{DEFAULT_BRANCH}}` failed, that breakage is this cycle's work and outranks the open-issue backlog — or, where the cause is outside anything this loop can change, file or link an issue for it and treat it as a blocking condition the same way Phase 8 treats a structurally-red anchor. While it stands, every PR inherits the red check, so the Phase 4 external anchor says nothing about the change under review. Before editing anything, run the Phase 3 verify commands once on an unmodified `{{DEFAULT_BRANCH}}` as a baseline, even when `{{DEFAULT_BRANCH}}` is green. A failure first seen after editing cannot be told apart from one the edit caused, and the cycle will spend turns proving it did not. **Close stale-open issues first.** Check whether any open issues were resolved by recent PRs but not yet closed. Cross-reference `git log` against open issue titles: From adcdd6b84634db11090cb18e5f30f9b50ca60f98 Mon Sep 17 00:00:00 2001 From: Daniel McCoy Stephenson Date: Sat, 14 Nov 2026 23:36:31 -0700 Subject: [PATCH 3/3] Record the PR number in the RESEARCH.md Implementations entry Co-Authored-By: Claude Opus 5.5 --- RESEARCH.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/RESEARCH.md b/RESEARCH.md index 7352bb2..7ca692f 100644 --- a/RESEARCH.md +++ b/RESEARCH.md @@ -52,7 +52,7 @@ Last reviewed: 2026-08-04. - PR #121 (Define the structurally-red-anchor path and require the anchor to have executed tests): shipped 2026-08-16. Observed effect: pending — needs N cycles of data. Extends the same argument as PR #54's "green CI is not verification when CI's scope excludes the changed files" to two further ways an anchor can carry no information: a required job that has *never* succeeded on the base branch (observed on a repo whose CI jobs have failed on empty credentials since February 2026, so every cycle terminated in a hand-off regardless of quality), and a build tool that reports success having executed zero tests (`:test NO-SOURCE` still prints `BUILD SUCCESSFUL`). Both were previously read as ordinary signals — the first as a blocking red, the second as a green — which is unanchored critique wearing an anchor's clothes. - PR #136 (Make the checkout form Phase 4's default Tests-fix revert and gate ladder rung 1): shipped 2026-08-27. Observed effect: pending — needs N cycles of data. Repairs the mechanism PR #54's entry above introduced. The prescribed `git stash push` revert ran against a tree Phase 3 had already committed and Phase 4 had already pushed, so it saved nothing, the check then passed against the still-fixed tree, and the no-op was naturally scored as a false-negative in the *test* — an anchor reporting a verdict it never measured, which is precisely the failure this finding warns about. The checkout form becomes the default because it is correct in the state Phase 4 actually finds, and the revert must now be confirmed to have changed the tree before either half is trusted. Separately, the ladder rung PR #67 added is gated against headless dispatch, where an interrupted session leaves the branch head undoing its own fix and the substitute anchor damages the thing it substitutes for. - PR #138 (Ground Phase 4's Docs score in a real pass): shipped 2026-08-29. Observed effect: pending — needs N cycles of data. The **Docs** rubric item named a check that Phase 7 did not perform until three phases later, so the only way to score it on time was from judgment — an anchor-shaped item with no anchor behind it, the same defect as PR #121's zero-executed-tests green, relocated from the signal to the schedule. The table pass now runs at Phase 4, where the score is taken, and Phase 7 is reduced to a re-check of whatever the review phases changed. Phase 7 is also required to record that it ran, since a clean sweep that leaves no artifact is indistinguishable from a skipped one — which is the auditability half of the same problem. -- PR #TBD (Check the default branch's current CI state and take a pre-edit baseline at Phase 1): shipped pending. Observed effect: pending — needs N cycles of data. PR #121 covered a required job that has *never* passed on the base branch, but not a job that passes routinely and is red at the moment. That case was seen in a generated instance whose `main` had failed its `Build` check for five days without any cycle noticing (#144). A PR branched off a red base inherits the red check, so the anchor says nothing about the change under review. Running the anchor once before editing means a later failure can be told apart from one the edit caused. +- PR #178 (Check the default branch's current CI state and take a pre-edit baseline at Phase 1): shipped pending merge. Observed effect: pending — needs N cycles of data. PR #121 covered a required job that has *never* passed on the base branch, but not a job that passes routinely and is red at the moment. That case was seen in a generated instance whose `main` had failed its `Build` check for five days without any cycle noticing (#144). A PR branched off a red base inherits the red check, so the anchor says nothing about the change under review. Running the anchor once before editing means a later failure can be told apart from one the edit caused. ---