Doctrine Self-Revision
Page 1 covered how doctrine gets read at the start of a session. This page covers the other direction: how it gets corrected. The rulebook here is not a spec written once and left to quietly go stale — it's treated as a living document that changes, in place, dated, every time real practice contradicts what it currently says.
This page is itself an instance of the cycle it describes — it exists because a genuine gap was caught mid-build and written up rather than quietly patched over.
What actually triggers a revision
Not a scheduled review. A revision gets written the same week a real incident happens — a mistake that actually occurred, an assumption that turned out false when checked against the real system, a judgment call that a closer look showed was wrong. Each entry keeps the concrete example: what was wrongly assumed, what was actually true, and a detection checklist so the same mistake is recognizable next time before it repeats.
Why this is a distinct mechanism, not just editing a file
The loop only closes if the correction actually reaches future sessions, not just this one. That's what makes it an architectural mechanism rather than routine documentation upkeep: a written correction feeds directly into Session Bootstrap, so the very next session — and every one after it — starts from the corrected understanding rather than the one that just failed. Without that read-back connection, "we fixed the doc" would mean nothing operationally.
Applies to this reference too
This Workflows section is not exempt from its own subject. If a future gap is found here — a real mechanism this page set doesn't cover, or a description that drifts from what the system actually does — the correct response is the same cycle: write it down, dated, with what changed, rather than quietly editing history.
The same principle, one level up
The SDLC, Paper 7: Papers Into Practice is the lived, specific case of this exact mechanism — not the doctrine files this page describes, but the Papers series itself, revising its own claims in public when practice contradicts them. Paper 7 is what this cycle looks like when the thing being corrected is an essay instead of a rulebook.
Updated July 12, 2026: added the cross-link to Paper 7, the concrete case this page had been describing only in the abstract.