Reference in New Issue
Block a user
Delete Branch "worktree-handoff-refresh"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Docs only —
HANDOFF.md. No plugin code changed in this repo this round, so no version bump (the handoff is a resume note, not a shipped artifact).Why this matters more than a status refresh
It corrects mislead-risk 4, which was half wrong in the direction that costs the next session the most time. Reproduced with both yq flavours installed locally:
check-routing.shdoes NOT false-green. Byte-identical output under kislyuk 3.4.3 and mikefarah v4.53.3, and it agrees with CI — expected, since oleks/claude-plugin-cluster#196 made its expression flavour-agnostic. The local/CI disagreement recorded in oleks/claude-plugin-cluster#252 is better explained by the known-flakylint-routing(oleks/claude-plugin-cluster#232); one of its two cited findings is the exactregistry-pins.txtstale claim documented there as appearing once and vanishing on rerun.generate-descriptions.shwas worse than reported — not a false green at all: it could not run under v4, exiting 3 withmikefarah v3? upgrade to v4while running on v4.The old text told a future reader to distrust both scripts and to add a "yq must be mikefarah v4" assert. Both instructions were wrong: one script was fine, and that assert would break the kislyuk path emmett actually uses.
Open items updated
claudecode-linterRecords explicitly that
CCL ⏭is the correct post-deploy output until the linter is actually packaged, so nobody "fixes" the skip token back into silence.The Unbanked-work section now names both remote branches with their heads, each verified server-side with
git ls-remotebefore its worktree was removed.Incidentally, this PR's own push printed
MDL ✅ CPV ✅— noCCLtoken — which is oleks/emmett#387 reproducing live in the act of documenting it.Round 2 touched no plugin code in this repo, so no version bump: HANDOFF.md is a resume note, not a shipped artifact. Corrects mislead-risk 4, which was half wrong in a way that would cost the next session real time. Reproduced with BOTH yq flavours installed locally: - check-routing.sh does NOT false-green. Byte-identical output under kislyuk 3.4.3 and mikefarah v4.53.3, and it agrees with CI. Expected, since #196 made its expression flavour-agnostic. The local/CI disagreement recorded in oleks/claude-plugin-cluster#252 is better explained by the known-flaky lint-routing (oleks/claude-plugin-cluster#232) -- one of its two cited findings is the exact registry-pins.txt stale claim documented there. - generate-descriptions.sh was WORSE than reported: not a false green, it could not run at all under v4, exiting 3 with "mikefarah v3? upgrade to v4" while running on v4. Open items updated: - oleks/claude-plugin-cluster#252 -> fixed, PR #253 (0.85.1), CI green, unmerged. - oleks/emmett#387 -> acceptance item 3 done in PR #388, deploy-pending; items 1-2 (package the linter) still open and belong on a remote builder. Records that CCL skip is the CORRECT post-deploy output until then, so nobody "fixes" the skip token back. - oleks/emmett#389 -> new: the silent-skip mode is systemic (SHC/HDL/YML/ TPL/CPV), not specific to claudecode-linter. Unbanked-work section now names both remote branches with their heads, both verified server-side via git ls-remote before their worktrees were removed. SC-003 (oleks/claude-plugin-cicd-insights#2) is unchanged and still the only item that requires a human.