Captured retroactively. This project shipped before the roadmap surface existed. The credit value is a post-hoc estimate per the methodology in Entry 017 and is subject to revision under G-038's calibration mechanism. This project lives outside the web app — it ships as a Claude skill in the OLN project root — but it's infrastructure in the same sense the corpora are: it's how new entries get authored consistently.
What shipped
~/ai-projects/oln/.claude/skills/decision-journal/SKILL.mddefining: when to trigger, dual-mode capture (append vs. session-notes), three entry shapes (journal / backlog / register), tier guidance, visibility guidance, sentinel pattern usage, ID conventions, full capture flow, what not to do.references/example-journal-entry.md— annotated journal entry using Entry 7 (brand architecture) as the worked example.references/example-register-entry.md— annotated phased register entry demonstrating the<!-- internal -->sentinel split.references/example-backlog-item.md— annotated backlog template.
Why
Without this skill, decision capture relies on the author remembering the conventions every time. The skill is what lets the corpora stay internally consistent across long stretches of work and across sessions where the prior context isn't loaded.