Migration Diagnose

migration-diagnose

The first step of the strangler-fig adoption track. Produce a structured report of every function, event handler, store action, or API call that is a candidate for wrapping as an acture command.

Read-only. No source files are modified.

Inputs

• The target codebase root.
• (Optional) An acture.config.json at the repo root with prefix (default app) and category hints.

Output

Write acture-output/diagnosis.md containing:

1. A one-line summary (state library, store count, candidate count).
2. A per-store table of actions.
3. A list of top-priority candidates (≥10 if available), each with proposed command id, current location, params shape, complexity, and priority.
4. A "skip" list (handlers we intentionally don't wrap — e.g. trivial local UI toggles).

Optionally also write acture-output/diagnosis.json if downstream tooling consumes it.

Steps

1. Detect state library

Search package.json dependencies and source code:

Record what you find. Most apps have one primary library plus React local state.

2. Locate stores / slices

For each detected library, find the store/slice files and enumerate the actions. For zustand, actions are methods inside the create((set, get) => ({...})) shape. For RTK, they are the reducers: keys of createSlice. Capture for each: file path, action name, parameter shape (from the TS signature if present), sync/async.

3. Find event handlers that mutate state

Grep for the high-signal patterns:

onClick={,

For each match, decide:

• Does it call a store action / API? → candidate
• Does it only flip local UI state (setIsOpen(...))? → skip
• Does it call multiple actions composed inline? → candidate (wrap the composition as one command)

4. Find API calls

fetch(, axios., .useQuery(, .useMutation(, .mutate(, `gql``. For each: endpoint or procedure name, input shape, sync/async, side-effecting?

5. Find pre-existing command-like patterns

Spot prior art the host already has — these are usually the highest-value candidates:

hotkeys(, tinykeys(, useHotkeys(            // keyboard
<CommandDialog, <KBarProvider, useKBar(     // palette
dispatch(, .execute(                         // central dispatchers

If the app already has a command palette or a hotkeys table, those entries are the canonical list of "things users invoke." Wrap them first.

6. Classify each candidate

For every item from steps 2–5 build an entry:

• Proposed id: {prefix}.{domain}.{action} (lowercase camelCase, e.g. app.todo.add).
• Current location: path/to/file.ts:LINE.
• Params: from the TS signature, or none.
• Complexity: simple (sync, ≤1 primitive param) / moderate (Zod schema or API call) / complex (multi-step, error paths, side effects).
• Priority (1–5):
  • 5: user-facing, called from ≥3 surfaces, core to the app's purpose.
  • 4: user-facing, called from 2 surfaces.
  • 3: user-facing, called from 1 surface.
  • 2: internal action that would benefit from telemetry / tests.
  • 1: utility unlikely to need multiple surfaces.

7. Write the report

acture-output/diagnosis.md should have these sections:

# Diagnosis

**State library:** zustand
**Stores found:** 1 (src/store.ts — todos)
**Total candidates:** 8
**Priority ≥ 4:** 5

## Stores

| Store | Actions | File |
|---|---|---|
| todos | addTodo, toggleTodo, removeTodo, clearDone | src/store.ts |

## Top candidates

| # | Proposed id | Location | Params | Complexity | Priority |
|---|---|---|---|---|---|
| 1 | app.todo.add | src/store.ts:42 | `{text: string}` | moderate | 5 |
| 2 | app.todo.toggle | src/store.ts:51 | `{id: string}` | simple | 4 |
| ... | | | | | |

## Skipped

- `setIsOpen` in src/Modal.tsx — local UI state only, not a candidate.

Validation

Before finishing, check:

• Every store file is enumerated.
• Every handler that calls a store action is captured as candidate or skip.
• Every proposed id matches ^[a-z][a-zA-Z0-9]*(\.[a-z][a-zA-Z0-9]*)*$ (acture's id regex).
• No source files were modified.
• The report is in acture-output/ and is readable as the input to migration-plan.

Hand-off

After this, run migration-plan to phase the candidates into a backlog. Do NOT jump to migration-wrap without a plan — the plan keeps the agent honest about Phase 1 / Phase 2 ordering and dependencies.

Source: thorwhalen/acture — distributed by TomeVault.