本文へ移動
cccskills
無料GitHub で公開

backend-authoritative-runtime-refactor

Refactor apps so backend owns runtime truth and UI only projects snapshots. Use when tools have split authority. NOT for cosmetic UI work or new features before drift is removed.

インストール方法を見る

含まれるファイル(8)

  • SKILL.md7.6 KB
  • affordance-scorecard.json804 B
  • CHANGELOG.md756 B
  • diagrams/01_flowchart_decision-points.md704 B
  • diagrams/INDEX.md141 B
  • references/architecture-patterns.md813 B
  • references/INDEX.md794 B
  • references/windags-lessons.md984 B

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Backend-Authoritative Runtime Refactor

Refactor only after you can say exactly where execution truth lives and how the UI reads it.

When to Use

  • The UI synthesizes graphs, phases, or provider choices that the backend does not understand.
  • Relaunch or reconnect changes visible runtime state because the frontend cached authority.
  • Old and new schemas, starter formats, or node types coexist in the active authoring path.
  • Approval, review, or downstream execution depends on frontend-only reconstruction.

NOT for

  • Cosmetic visual cleanup with no change to runtime ownership.
  • Isolated component refactors where the backend contract is already authoritative.
  • Shipping new business features before deleting the split-authority seam.
  • Pure backend performance work when the architecture already has one source of truth.

Decision Points

flowchart TD
  A[Workflow refactor request] --> B{Can the UI still invent runtime truth?}
  B -->|No| C[Use a narrower domain or UI skill]
  B -->|Yes| D[Inventory command, query, and event seams]
  D --> E{Can every user action map to command, query, or event?}
  E -->|No| F[Freeze the contract before feature work]
  E -->|Yes| G[Split projection state from backend-owned state]
  G --> H{Are legacy compat paths still active?}
  H -->|Yes| I[Move them to migration edges or delete them]
  H -->|No| J[Add evaluators, replay-safe startup queries, and event subscriptions]
  I --> J
  J --> K[Validate reconnection, approval, and downstream readiness]

Use this routing model first:

  • If the UI cannot be described as command -> backend transition -> query or event projection, the seam is still wrong.
  • If a "temporary compatibility" path sits in the hot path, treat it as refactor debt, not a feature.
  • If restart or reconnect cannot reconstruct visible state from backend truth alone, the refactor is incomplete.

Working Method

1. Inventory the split brain

Find all places where the UI:

  • generates or mutates execution truth
  • invents phases that the backend does not know about
  • simulates execution in active code paths
  • preserves both old and new schemas in the same authoring path
  • mixes abstract routing tiers with exact provider IDs

2. Freeze the contract

Define three classes only:

  • Commands: start, stop, approve, resume, mutate
  • Queries: startup state, review payload, execution snapshot, provider health
  • Events: authoritative transitions, evaluator verdicts, gate status, replay markers

If a UI action cannot be expressed as one of these, the architecture is still muddy.

3. Normalize state ownership

Split stores into:

  • contract input state
  • backend snapshot projection state
  • UI layout state

Do not let any store own both runtime truth and layout concerns.

4. Delete active compatibility seams

Prefer deletion over "supporting both for now."

Typical deletions:

  • singular and plural schema compatibility in active authoring
  • browser or demo simulation in the active desktop runtime
  • deprecated node types in the main path
  • duplicate review summaries
  • multiple starter or import formats

5. Add evaluators before trusting the refactor

Every implementation node should have:

  • a local-output evaluator for correctness
  • a downstream-readiness evaluator for contract safety
  • a restart or reconnect path that rehydrates from backend truth, not frontend guesses

Failure Modes

Projection masquerading as authority

Symptom: the UI looks "read only" but quietly assembles phase, graph, or provider truth locally.

Fix: move that logic into backend commands or queries and treat the frontend as a projector.

Compatibility seam calcification

Symptom: "support both for now" remains in the hot path for weeks and becomes the real architecture.

Fix: push compatibility to one-way migration edges with an explicit deletion date or remove it now.

Approval without inspectability

Symptom: a human can approve work they cannot inspect because the review payload is partial or frontend synthesized.

Fix: backend must emit the exact review payload; the UI only renders it.

Replay-hostile startup

Symptom: relaunching the app changes visible state or loses gate position because the frontend cached authority.

Fix: startup must be a query against backend truth plus event replay, not store reconstruction.

Shibboleth

If relaunching the app changes visible runtime state without any new backend event, authority is still split.

Worked Example

Tauri workflow editor with frontend-owned phase logic

Problem:

  • React stores authored the graph, inferred phase state, and normalized provider tiers.
  • Backend execution used a different contract, so review and restart flows drifted.

Refactor path:

  1. Inventory every place the UI invented execution truth.
  2. Collapse runtime interactions to commands, queries, and events.
  3. Move provider normalization and phase transitions behind backend commands.
  4. Delete starter-format unions from the active authoring path.
  5. Add evaluator gates so downstream work only proceeds from backend-validated outputs.

Success condition:

  • On restart, the UI reconstructs the same visible state from backend snapshot plus events.
  • Approval screens render a backend-emitted review payload, not a React recomposition.

Fork Guidance

Fork when the work separates cleanly:

  • Contract lane: inventory split-authority seams and define commands, queries, and events.
  • Deletion lane: remove compatibility layers, deprecated node types, and starter-format unions.
  • Verification lane: add evaluator gates, replay tests, and restart or reconnect coverage.

Keep final ownership decisions in the parent lane so one actor decides what the backend truly owns.

Quality Gates

  • The UI cannot fabricate graph, phase, provider, or gate truth locally.
  • Every user action maps to a command, query, or event.
  • Runtime state survives relaunch and reconnect from backend truth alone.
  • Review and approval payloads are emitted by the backend and rendered by the UI.
  • Compatibility logic lives only on migration edges, not in the active path.
  • Downstream execution stops on upstream contract failure.
  • Exact provider IDs and model choices are visible at the contract boundary.
  • Tests cover restart, reconnect, approval, and evaluator failure paths.

Reference Map

  • references/architecture-patterns.md — backend-owned workflow contract patterns and state-ownership splits.
  • references/windags-lessons.md — concrete lessons from a real split-authority runtime migration.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Expert in 2000s-era music visualization (Milkdrop, AVS, Geiss) and modern WebGL implementations. Specializes in Butterchurn integration, Web Audio API AnalyserNode FFT data, GLSL shaders for audio-reactive visuals, and psychedelic generative art. Activate on "Milkdrop", "music visualization", "WebGL visualizer", "Butterchurn", "audio reactive", "FFT visualization", "spectrum analyzer". NOT for simple bar charts/waveforms (use basic canvas), video editing, or non-audio visuals.

日本語の概要は準備中です。原文の説明を表示しています。

curiositech/port-daddy22026年10月8日 更新

Expert legal research agent for finding and scraping expungement data state by state. Knows authoritative sources, URL patterns, Firecrawl configuration, and 2026 legal landscape.

日本語の概要は準備中です。原文の説明を表示しています。

curiositech/port-daddy22026年10月8日 更新

Expert in 3D computer vision labeling tools, workflows, and AI-assisted annotation for LiDAR, point clouds, and sensor fusion. Covers SAM4D/Point-SAM, human-in-the-loop architectures, and vertical-specific training strategies. Activate on '3D labeling', 'point cloud annotation', 'LiDAR labeling', 'SAM 3D', 'SAM4D', 'sensor fusion annotation', '3D bounding box', 'semantic segmentation point cloud'. NOT for 2D image labeling (use clip-aware-embeddings), general ML training (use ml-engineer), video annotation without 3D (use computer-vision-pipeline), or VLM prompt engineering (use prompt-engineer).

日本語の概要は準備中です。原文の説明を表示しています。

curiositech/port-daddy22026年10月8日 更新

Apply crisis decision-making research to agent routing, uncertainty triage, and coordination failure analysis in time-pressured systems. Use when diagnosing handoff failures, analytical paralysis, or expert judgment under incomplete information. NOT for routine coding, simple CRUD design, or static single-agent tasks with complete information.

日本語の概要は準備中です。原文の説明を表示しています。

curiositech/port-daddy22026年10月8日 更新

Use for insight, reframing, contradiction, impasse, and anomaly-driven problem solving when execution effort no longer helps. NOT for routine optimization, error correction, or well-specified tasks with known solution paths.

日本語の概要は準備中です。原文の説明を表示しています。

curiositech/port-daddy22026年10月8日 更新

Apply cognitive task analysis to expert work that depends on perceptual cues, branching judgment, and recurring monitoring loops. Use when decomposing expert capability into agent structure, simulation design, or validation interviews. NOT for ordinary step-by-step SOP capture or simple pipelines with no tacit cue layer.

日本語の概要は準備中です。原文の説明を表示しています。

curiositech/port-daddy22026年10月8日 更新

curiositech のスキルをすべて見る

このスキルの問題を報告する