# Analyzer Runtime

Analyzers are thin diagnostic layers over the target project's ESLint setup.
They reuse the project's parser, resolver, plugins, ignores, and framework
configuration, then append topic-specific rules for the diagnostic run.

## Runtime Pipeline

```mermaid
flowchart TD
  command[Analyze command] --> parse[Parse topic and options]
  parse --> load[Load target eslint.config.ts]
  load --> validate[Validate productive-eslint marker]
  validate --> topic[Select analyzer]
  topic --> clone[Clone FlatConfigComposer]
  clone --> overlay[Append analyzer overlay]
  overlay --> eslint[Run ESLint]
  eslint --> findings[Map lint messages to findings]
  findings --> score[Score and classify]
  score --> summarize[Summarize by file]
  summarize --> markdown[Render Markdown]
```

## Config Loading

The loader temporarily switches `process.cwd()` to the selected `--cwd` while it
imports the target `eslint.config.ts` or `eslint.config.mts`.

This matters for project-local auto-detection. For example, Vue or RxJS support
should be detected from the target package root, not from the directory where
the CLI process happened to start.

```mermaid
flowchart LR
  cliCwd[CLI process cwd] --> save[Save cwd]
  save --> target[Switch to --cwd]
  target --> importConfig[Import eslint.config.ts]
  importConfig --> restore[Restore previous cwd]
```

## Report Model

Every analyzer returns the same structured report shape:

- title;
- scanned file count;
- finding count;
- summary lines;
- ranked file summaries;
- suggested order;
- next step.

Markdown is rendered from that structured model.

```mermaid
classDiagram
  class AnalyzerReport {
    string title
    number fileCount
    number findingsCount
    string[] summaryLines
    FileSummary[] summaries
    string[] suggestedOrder
    string nextStep
  }

  class FileSummary {
    string file
    number score
    string[] labels
    string[] reasons
    Finding[] findings
  }

  class Finding {
    string file
    string ruleId
    string severity
    number score
    string[] labels
    string[] reasons
  }

  AnalyzerReport --> FileSummary
  FileSummary --> Finding
```

## Analyzer Responsibility

Analyzers should:

- reuse ESLint as the canonical finding source when possible;
- add focused rule overlays for the selected topic;
- use analyzer-local plugin namespaces when a diagnostic rule must run
  independently from the target project's optional plugin setup;
- classify and score findings;
- group findings into review-friendly hotspots;
- keep output advisory rather than punitive.

Analyzers should not:

- replace the target project's ESLint configuration;
- run implicitly during normal coding tasks;
- treat findings as process failures;
- implement a second parser or resolver model when ESLint already provides one.