DR-Next Snapshot Replay
Consumer contract for stage snapshots — /dr-next and /dr-orchestrate read datarim/snapshots/{TASK-ID}.snapshot.md first, emit replay-prompt with CTA + bilingual autonomy reminder + done before: body.
Overview
DR-Next Snapshot Replay is the consumer contract for per-task stage snapshots — introduced in v2.13.0 (TUNE-0254). /dr-next and /dr-orchestrate, after task resolution, read datarim/snapshots/{TASK-ID}.snapshot.md first — before task-description, init-task, activeContext, PRD or plan — and emit a replay-prompt that lets the agent pick up where the previous stage left off.
Replay-prompt template (canonical)
<recommended-CTA>
ищи способ исследовать все проблемы и ответить на все вопросы самостоятельно. выполняй за оператора все необходимые шаги и требования, которые можешь сделать сам. не создавай FU подзадачи, если они уровня 1, а решай их в этом же цикле до полного решения.
Find a way to investigate all problems and answer all questions yourself. Perform on behalf of the operator every step and requirement you can do yourself. Do not spawn FU sub-tasks for Level-1 work — resolve them in the same cycle to completion.
done before:
<snapshot body>
Why bilingual
The autonomy reminder is intentionally duplicated in Russian (primary, per project mandate) and English (dub guard). When the agent runs under Codex CLI or an EN-locale agent the Russian text may be parsed less reliably; the EN dub guarantees execution regardless of runtime locale.
CTA selection heuristic
The recommended_next in the snapshot was emitted by the previous stage according to the standard cta-format.md § Authoring Rules — it is already correct. The replay skill documents the rationale for preferring that option, not a re-derivation. Principle: maximise the quality delta of the solution. For L3+ tasks with few completed checks, verification-style commands (/dr-verify, /dr-qa, /dr-design) are preferred over /dr-do; when checks are saturated, the heuristic transitions to implementation or archiving. Three worked examples cover L3+ with few checks → /dr-verify, L3+ saturated → /dr-do, L1/L2 after /dr-do → /dr-archive.
Fallback when no snapshot exists
If the snapshot file is missing or its frontmatter fails validation (dev-tools/check-stage-snapshot-on-exit.sh --validate-frontmatter --task <ID> exit ≠ 0), the consumer silently falls through to the legacy Read pipeline — no warning lines, no behaviour change for first-time /dr-next invocations.
Orchestrator integration
In /dr-orchestrate Phase 2 cycle the snapshot read happens before subagent_resolver.sh. recommended_next is passed to the resolver as --hint <command> — the resolver can still return a different command (snapshot is a hint, not a constraint).
Source
TUNE-0254, v2.13.0, 2026-05-21. Consumer touchpoints: commands/dr-next.md Step 2.5, plugins/dr-orchestrate/commands/dr-orchestrate.md Snapshot-First Resume. Producer contract sibling: skills/stage-snapshot-writer.md.