Finish One World

This sprint gave later workers a shared source of truth before they changed the format. Its scope was deliberately document-only: point the agent orientation path at the pattern material and state which document governs format changes. The gate and review passed cleanly, which fits a unit whose risk was ambiguity rather than runtime behavior. That made the following declaration sprints start from a named authority instead of scattered examples.

This was the first production dispatch removal. The worker had to add an optional stage declaration, preserve existing bundled output, and keep the future-work-list behavior structural rather than prompt-owned. The sprint stayed green, suggesting the boundary around display classification and parser tests was small enough for a single fresh context. It left later role work with a concrete defaulting rule instead of a prompt-name table.

This sprint consumed the stage defaulting contract and narrowed what role strings are allowed to mean. Its job was to separate closed parsing vocabulary from author identity, while keeping the detector strict. The run facts show the ordinary green path, with no repair or recovery attached to this unit. That is consistent with a small semantic audit backed by parser and interpreter tests.

This sprint removed the review recovery case that depended on a bundled channel name. The brief framed a precise replacement: channels declare that their prose uses review-verdict recovery format, and recovery follows that declaration. The run record has no red gate here, so the tricky part stayed inside the channel contract rather than spilling into later control flow. Production and interpreter coverage were part of done, which kept the recovery from becoming a doc-only claim.

This sprint closed the first milestone after the declaration changes landed. It was asked to make the spec, manual, and enforcement posture agree, including display stages, role defaults, future-work-list behavior, and declared verdict recovery. The sprint passed through the normal checks without repair. Its build boundary was editorial and enforcement alignment: no new hidden branch could be accepted unless the written rule explained it.

This sprint established the declaration route for standing loops before deleting the old state. Its brief was broad but cohesive: one standing iteration agent, typed inputs, counters, conditions, channels, resume, and validation behavior. The run facts show a replan on this sprint, which marks the boundary where partial conversion needed a sharper follow-up. That let the next sprint focus on completion and deletion rather than rediscovering the whole contract.

This sprint took the partial standing-loop declaration work over the deletion line. It had to bind the remaining loop inputs through declared names, move quiescence and done handling into general counters and channels, and remove the old production-world state. The run facts show the unit passing without a recovery stage. Because the prior sprint exposed the missing pieces, this worker could treat the task as completion of a known contract.

This sprint performed the equivalent conversion for goal cycles after standing-loop removal had proven the pattern. Its brief separated declared pattern state from the goal-check service that may remain in the engine. The unit stayed green, with no recorded repair, so the risk was contained in specializing the right engine authority and leaving prompt, test, ceiling, snapshot, and resume data in declared homes. That made the closure sprint mostly an audit of remaining state.

This sprint was the final cleanup pass for the runtime-removal milestone. The ledger note says the work was already satisfied and the gate was green, so the worker recorded completion after rechecking the closure criteria. That makes this a thin but real unit: it existed to prove no remaining goal-cycle state or enforcement debt had survived the prior conversions. It also kept the milestone from ending on an implicit assumption.

This sprint protected older pinned runs after the runtime contracts were removed. The brief kept the work at the resume boundary: stamp current pins, check the marker before parsing frozen closures, and report older pins truthfully. The gate stayed green, which matters because the sprint touched failure timing rather than normal execution. It exposed a contract the later source-parity and docs sprints could reference.

This sprint aligned shipped map-spec variants around declared plan-source handling. Its scope was small but compatibility-sensitive because changing the declared source can change run identity. The unit passed cleanly and documented the identity implication in engineering docs rather than release material. It gave the later authored coverage a consistent read-side surface to assert against.

This sprint was coverage for a historical production-world gap rather than a new user command. It asked an authored fan-out pattern to run a per-unit body with review under its own names, away from shipped prompt ids and channel spellings. The run facts show a clean gate and review, so the acid test did not uncover another production dispatch branch. Its boundary was proving fan-out review behavior where interpreter-only tests had not been enough.

This sprint composed the strongest authored-pattern acid test. Instead of checking isolated features, it had to run a custom pattern through plan source recovery, roster titles, per-entry completion, review recovery, and replanning under its own names. The clean run record suggests the prior declaration work had already removed the blocking special cases. The sprint's value is the end-to-end boundary: a fresh worker could hold one fixture while the fixture exercised the whole shape.

This sprint consumed the documentation drift found after runtime removal. Its brief was a focused reconciliation pass over the manual, spec, compatibility marker, plan-source parity, and enforcement posture. It stayed green, which fits a unit whose risk was stale guidance rather than runtime failure. The placement after executing coverage meant the docs described behavior already exercised by the engine.

This final sprint checked the heavy-tier finding and the remaining enforcement posture after the behavioral and documentation work had landed. The ledger records it as already satisfied with a green gate, so it is intentionally a closure unit. It did not need a new repair loop; it needed evidence that temporary production-world debt had not survived the earlier sprints. That made the run finish with a checked boundary instead of an open heavy-test note.