CHANGELOG.md:214-216 claims 'lint over graphs/*.yaml is unchanged from the pre-edit baseline on both runtimes — claude 4 warnings, codex 11 warnings plus adr-driven-dev's pre-existing exit 1', without saying how to reproduce it. A reader who reaches for the obvious form gets an error: go run ./cmd/oh-my-graph lint graphs/backlog-batch.yaml --runtime codex exits 1 with 'oh-my-graph: lint: unexpected argument "--runtime" (usage: oh-my-graph lint <graph.yaml>)' (measured 2026-08-22 on branch lane-b2b). The flag is global and must precede the subcommand — cmd/oh-my-graph/lint.go:21 rejects a dash-prefixed positional and :27-28 rejects any second argument, while the runtime arrives from the global dispatch at cmd/oh-my-graph/main.go:195. The working form is go run ./cmd/oh-my-graph --runtime codex lint graphs/backlog-batch.yaml. The CLI half of this is already tracked as backlog item 30 (a usage line that never says where the flag goes); what is new here is that a CHANGELOG figure was published with no retraceable command next to it. The fix wanted is the reproduction command in the CHANGELOG entry, not a code change.
CHANGELOG.md:214-216 claims 'lint over graphs/*.yaml is unchanged from the pre-edit baseline on both runtimes — claude 4 warnings, codex 11 warnings plus adr-driven-dev's pre-existing exit 1', without saying how to reproduce it. A reader who reaches for the obvious form gets an error: go run ./cmd/oh-my-graph lint graphs/backlog-batch.yaml --runtime codex exits 1 with 'oh-my-graph: lint: unexpected argument "--runtime" (usage: oh-my-graph lint <graph.yaml>)' (measured 2026-08-22 on branch lane-b2b). The flag is global and must precede the subcommand — cmd/oh-my-graph/lint.go:21 rejects a dash-prefixed positional and :27-28 rejects any second argument, while the runtime arrives from the global dispatch at cmd/oh-my-graph/main.go:195. The working form is go run ./cmd/oh-my-graph --runtime codex lint graphs/backlog-batch.yaml. The CLI half of this is already tracked as backlog item 30 (a usage line that never says where the flag goes); what is new here is that a CHANGELOG figure was published with no retraceable command next to it. The fix wanted is the reproduction command in the CHANGELOG entry, not a code change.