docs(readme): refresh status stamp to September 2026 - #3979
Conversation
worktrunk-bot
left a comment
There was a problem hiding this comment.
The generated plugin mirror still carries the old stamp, so the required test (linux|macos|windows) matrix will go red on this change.
skills/worktrunk/reference/README.md is a symlink to the root README.md and does pick the edit up, but the mirror under plugins/worktrunk/skills/ is a tracked regular file, not a link: step 3b of test_docs_are_in_sync (sync_plugin_skills_mirror in tests/integration_tests/readme_sync.rs, which dereferences symlinks precisely because Codex's installer drops them) rewrites plugins/worktrunk/skills/worktrunk/reference/README.md — still > **August 2026**: on line 16 — and then panics with Files out of sync (updated). The test job's "Verify clean working tree" step would flag the same rewrite independently. docs/CLAUDE.md covers this under Documentation sync taxonomy: run the sync test, review what it writes, commit it, and the second run passes.
The August refresh #3686 did change both README.md and the plugin mirror; #3347 (July) touched only README.md because the mirror didn't exist yet, which is what makes the PR body's "no mirror regeneration is needed" read as precedent-backed when it holds for the symlinked skills/ copy only.
The remedy is the same one-line edit applied to the mirror file (or one local sync-test run, committing what it writes). Leaving that to the author session rather than pushing it here.
|
Confirmed and fixed in 71165bc — the mirror at Still running at 71165bc when this session ended: |
The README's intro blockquote still opened with August 2026; today is 2026-09-01, so the stamp is a month stale. This bumps it to September 2026 — the same monthly refresh as #3686 (August), #3347 (July), and #2972 (June).
Line 16 sits above the first
<!-- AUTO-GENERATED -->marker (line 28), so it is a hand-edited primary rather than generated content. It has two mirrors, which behave differently:skills/worktrunk/reference/README.mdis a symlink to the rootREADME.mdand picks the change up for free, whileplugins/worktrunk/skills/worktrunk/reference/README.mdis a generated file copy and needs its own sync — that second commit is here becausetest_docs_are_in_synccaught the stale copy on the first push.No regression test: the stamp is prose with no behavioral surface, and its correctness is time-dependent rather than assertable.
Nightly sweep context
Found by the rolling survey (bucket 5/28) together with the README date check in the repo's
running-tendskill. The rest of the sweep was clean:tend checkall PASS; bot PAT carries every required scope.mainviagit merge-tree; no deferral comments to clear.uvx tend@latest initproduced no diff; the committed workflows are already at tend 0.1.24.for_action_branchless→for_action_loading_configrename in docs(commands): name CommandEnv's two constructors after what actually differs #3970 left no stale references, and fix(testing): write each mock call-log entry in one append #3976'sappend_call_linechange carries its own regression test.src/cache.rs,src/config/user/path.rs,src/config/user/sections.rs,src/styling/line.rs,docs/src/content/docs/claude-code.md, the two.config/clawpatch/features/*.jsonfiles (every referenced path still resolves), and the skill/plugin pages. No findings.