ln-26-architecture-migration-planner

Architecture Migration Planner

Goal: Plan a safe transition from evidenced current architecture to an explicit target, with reversibility limits and recovery actions. Change only the approved migration document; do not execute migrations, edit product code, build a generic task plan, approve delivery, or hide irreversible steps.

Execution contract: The checklist defines completion. Track each item internally as PENDING, PROVEN with evidence, CLEARED with evidence its condition is absent, or UNPROVEN with a gap; reading, delegation, tool failure, a zero exit status, or a self-reported success is not proof; only the observed outcome is. Reconcile after each section. Before returning, resolve all PENDING, count only PROVEN and CLEARED, and apply verdict and approval rules to every gap. Preserve intent, scope, and existing authorization. Continue authorized work; ask only for consequential unresolved choices or required external approval. When no one can answer during the run, state the exact question and apply the skill's verdict for the remaining gap instead of waiting or guessing. Scale depth to material risk without skipping checks. Preserve dependency and safety order; otherwise choose an appropriate verification method. Accept equivalent user or repository evidence; no other skill, named artifact, or complete lifecycle is required. Preserve source requirement and decision IDs. Bind reused evidence to relevant source versions, dirty changes, configuration, and environment; invalidate only affected claims. On continuation, reconcile task, authorization, current state, and unresolved evidence. For long work, return a compact continuation record or update an already authorized artifact; read-only skills do not persist it. Distinguish artifact readiness, verified behavior, and external-action authority. Prepare authorized work before required approval. If blocked by an instruction, cite its exact source and unresolved boundary; do not invent approval gates from caution.

Tool Routing

Need	Preferred capability	Fallback
Current and target states	Repository evidence plus approved architecture artifacts	Explicit user-provided states with limitations
Consumers and compatibility	Language intelligence, schema tools, config search, telemetry, and direct inspection	Conservative inventory marked UNVERIFIED
Data scale and runtime risk	Production metrics, checked-in reports, migrations, and workload evidence	Ranges with validation gates before execution
External migration constraints	Official vendor migration and compatibility guidance	Mark dependent phases BLOCKED or UNVERIFIED
Document mutation	Minimal patch to the approved migration-plan artifact	Return BLOCKED when authority or path is unclear

A migration phase must leave the system in a supported state. Additive and reversible steps precede cutover; destructive cleanup follows verified zero use.

Artifact Rules

• Reuse a clear architecture migration document; otherwise use docs/architecture/migration-plan.md.
• Treat current-state, target-design, baseline, decisions, diagrams, interfaces, and telemetry as optional shared evidence.
• Keep the plan architectural: phases, compatibility, data, topology, gates, rollback, ownership, and removal.
• Leave file-level implementation tasks to downstream planning.
• Never describe rollback as "revert" when data or external effects are not reversible.
• Give every destructive step explicit approval, backup, restoration, and zero-consumer evidence requirements.

Checklist

1. Establish the Transition Contract

• Resolve migration scope, business outcome, current state, target state, non-goals, deadline or horizon, owners, and approved destination.
• Read repository instructions, Git state, relevant architecture artifacts, migrations, deployment configuration, and compatibility policies.
• Verify that current and target states are specific enough to compute a gap; return BLOCKED rather than invent either state.
• Identify protected user journeys, invariants, SLOs, recovery objectives, compliance duties, and change windows.
• Keep the run read-only except for the approved migration document.

2. Build the Gap and Dependency Map

• Inventory affected modules, deployables, data stores, schemas, APIs, events, configuration, infrastructure, and operational procedures.
• Inventory internal and external consumers, owners, versions, traffic, data volume, and evidence quality.
• Map current-to-target changes in boundaries, ownership, contracts, data, runtime topology, observability, and failure behavior.
• Identify shared mutable resources, sequencing dependencies, long-running work, mixed-version windows, and irreversible effects.
• Record unknown consumers or usage as migration risks; absence of search results is not zero usage.

3. Design Compatibility and Data Safety

• Define old/new contract compatibility, version negotiation, adapters, dual-read or dual-write behavior, and deprecation policy where relevant.
• Use expand/migrate/contract when mixed versions or online migration require coexistence; separate additive and destructive releases in that case. For an approved offline atomic transition, document the equivalent compatibility, recovery, and downtime controls.
• Define backfill selection, batching, throttling, idempotency, checkpoints, retries, reconciliation, and correctness oracle.
• Define source of truth during coexistence, concurrent-write conflict handling, and detection, retry, and reconciliation of partial dual-write failure; do not assume two writes are atomic.
• Define backup, restore, RPO/RTO impact, privacy, retention, and audit evidence for data movement.

4. Build Reversible Phases

• Define preparation, shadow or coexistence, progressive migration, cutover, stabilization, and old-path removal as independently verifiable phases.
• For every phase, state prerequisites, changed architecture state, owner, entry gate, observable success, abort condition, rollback or roll-forward action, and exit evidence.
• Define feature flags, routing controls, canary cohorts, rate limits, maintenance windows, and blast-radius controls where justified.
• Define metrics, logs, traces, reconciliation reports, dashboards, alerts, and SLO gates needed before traffic or data movement.
• Keep old and new versions interoperable through realistic deployment ordering and rollback windows.

5. Plan Cutover and Removal

• Define go/no-go authority, communication, freeze conditions, exact cutover control, and immediate verification.
• Define rollback boundaries separately for code, configuration, traffic, schema, and migrated data, including points of no return and roll-forward recovery when rollback would lose accepted writes.
• Before destructive removal, require usage evidence over a window covering relevant schedules and offline consumers, migrated consumers, applicable retention expiry, and explicit approval. A short zero-traffic sample cannot establish zero use.
• List old code paths, contracts, flags, adapters, jobs, data, infrastructure, dashboards, and documentation to remove.
• Define post-cutover observation period, ownership handoff, incident response, and closure evidence.

6. Write and Report

• Write transition summary, state gap, dependencies, compatibility, data plan, phased sequence, gates, observability, rollback, removal, owners, assumptions, and open decisions.
• Link shared artifacts by stable repository path or title without requiring a particular workflow.
• Re-read every phase for unsupported zero-downtime, zero-loss, consumer, capacity, or reversibility claims.
• Link each transition unit to the protected requirement, current/target boundary, entry evidence, observed success, and irreversible point; identify dependent plans affected by a migration change.
• Apply the verdict to the plan artifact: READY only when no unresolved decision changes planned data semantics, compatibility, phase order, or recovery; INCOMPLETE for a useful plan with any such decision or missing gate/rollback rule. Assigning an owner or deferring a decision to a phase gate does not resolve it. Use BLOCKED only when scope, authority, essential current/target evidence, or safe document creation cannot be established. Pending live verification alone is an execution prerequisite, not a reason to block a useful plan.

Self-Check

• Reconcile before returning. Check item-level evidence, requirement coverage, contradictions, scope, verdict, and applicable cleanup. Correct the report or authorized artifacts. Reuse valid evidence; do not automatically rescan the repository or rerun successful commands. Repeat checks only for relevant changes, failures, or unresolved evidence. Disclose remaining gaps.

Output Contract

Report in the user's language, in this order; label all five fields and state each fact once. Use controlled plain language: one fact per sentence, usually under 20 words, active voice, and one term per concept, with no synonyms for verdicts, IDs, or states. Small results may use one line per field; omit empty tables and do not copy linked artifacts:

1. Result: The exact skill-specific verdict token first, then the supported outcome.
2. Scope: Reviewed/changed scope, exclusions, baseline, and material assumptions.
3. Evidence: Skill-specific fields below; distinguish facts, inferences, and unverified claims. Link artifacts; use tables when useful.
4. Verification: Checks/results, unavailable evidence, and applicable cleanup/external state.
5. Completion: Checklist: X/Y complete; Incomplete: None or each UNPROVEN item's reason, outcome impact, and exact next action; residual risks and required decisions.

Skill-specific evidence: Artifact path; current/target gap, consumers, dependencies, and protected invariants. Summarize phases with entry gates, changes, success evidence, abort conditions, rollback/roll-forward; data/coexistence/reconciliation/cutover/removal proof; and open decisions that affect safety, ordering, or reversibility. When there are more than a few phases and the host renders Markdown diagrams, add one Mermaid phase-and-gate diagram; keep the text complete without it.