@gobing-ai/ts-rule-engine 0.2.7 → 0.2.9

package/schemas/preset.schema.json

@@ -0,0 +1,54 @@
+{
+    "$schema": "https://json-schema.org/draft/2020-12/schema",
+    "$id": "https://raw.githubusercontent.com/gobing-ai/ts-libs/main/packages/rule-engine/schemas/preset.schema.json",
+    "title": "@gobing-ai/ts-rule-engine Preset",
+    "type": "object",
+    "additionalProperties": false,
+    "required": ["name"],
+    "properties": {
+        "$schema": { "type": "string" },
+        "name": { "type": "string", "minLength": 1 },
+        "extends": { "type": "array", "items": { "type": "string" } },
+        "disable": { "type": "array", "items": { "type": "string" } },
+        "overrides": {
+            "type": "object",
+            "additionalProperties": {
+                "type": "object",
+                "additionalProperties": false,
+                "properties": {
+                    "fix": {
+                        "type": "object",
+                        "additionalProperties": false,
+                        "properties": {
+                            "mode": { "enum": ["none", "suggest", "auto"] }
+                        }
+                    }
+                }
+            }
+        },
+        "extensions": {
+            "type": "object",
+            "additionalProperties": false,
+            "properties": {
+                "resolvers": { "type": "array", "items": { "$ref": "#/$defs/relativeExtensionPath" } },
+                "evaluators": { "type": "array", "items": { "$ref": "#/$defs/relativeExtensionPath" } },
+                "fixers": { "type": "array", "items": { "$ref": "#/$defs/relativeExtensionPath" } },
+                "formatters": { "type": "array", "items": { "$ref": "#/$defs/relativeExtensionPath" } }
+            }
+        }
+    },
+    "$defs": {
+        "relativeExtensionPath": {
+            "type": "string",
+            "minLength": 1,
+            "description": "Relative module path; absolute paths and '..' traversal are forbidden.",
+            "not": {
+                "anyOf": [
+                    { "pattern": "^[/\\\\]" },
+                    { "pattern": "^[A-Za-z]:[/\\\\]" },
+                    { "pattern": "(^|[/\\\\])\\.\\.([/\\\\]|$)" }
+                ]
+            }
+        }
+    }
+}

package/schemas/rule-file.schema.json

@@ -0,0 +1,49 @@
+{
+    "$schema": "https://json-schema.org/draft/2020-12/schema",
+    "$id": "https://raw.githubusercontent.com/gobing-ai/ts-libs/main/packages/rule-engine/schemas/rule-file.schema.json",
+    "title": "@gobing-ai/ts-rule-engine Rule File",
+    "type": "object",
+    "additionalProperties": false,
+    "required": ["rules"],
+    "properties": {
+        "$schema": { "type": "string" },
+        "include": { "type": "array", "items": { "type": "string" } },
+        "exclude": { "type": "array", "items": { "type": "string" } },
+        "severity": { "enum": ["error", "warning", "info"] },
+        "rules": { "type": "array", "items": { "$ref": "#/$defs/rule" } }
+    },
+    "$defs": {
+        "rule": {
+            "type": "object",
+            "additionalProperties": false,
+            "required": ["id", "evaluator"],
+            "properties": {
+                "id": { "type": "string" },
+                "description": { "type": "string" },
+                "enabled": { "type": "boolean" },
+                "severity": { "enum": ["error", "warning", "info"] },
+                "include": { "type": "array", "items": { "type": "string" } },
+                "exclude": { "type": "array", "items": { "type": "string" } },
+                "evaluator": {
+                    "type": "object",
+                    "additionalProperties": false,
+                    "required": ["type"],
+                    "properties": {
+                        "type": { "type": "string" },
+                        "config": { "type": "object", "additionalProperties": true }
+                    }
+                },
+                "fix": { "$ref": "#/$defs/fix" }
+            }
+        },
+        "fix": {
+            "type": "object",
+            "additionalProperties": false,
+            "properties": {
+                "mode": { "enum": ["none", "suggest", "auto"] },
+                "replacement": { "type": "string" },
+                "params": { "type": "object", "additionalProperties": true }
+            }
+        }
+    }
+}

package/src/config/extensions.ts

@@ -24,6 +24,8 @@ export interface LoadExtensionsOptions {
     allowExtensions?: boolean;
     /** Optional sink for non-fatal warnings (e.g. built-in overrides). */
     logger?: { warn: (message: string) => void };
+    /** Optional module loader seam for tests or embedders with custom import policy. */
+    moduleLoader?: (absPath: string) => Promise<Record<string, unknown>>;
 }
 
 /** Host registries that can receive extension capabilities (fixers live on the engine, not the host). */
@@ -34,21 +36,22 @@ const HOST_REGISTRY_BY_KIND: Partial<Record<ExtensionKind, 'resolvers' | 'evalua
 };
 
 /**
- * Collect extension refs declared by a preset's `extensions` block.
+ * Collect extension refs declared by a preset's or rule file's `extensions` block.
  *
- * Paths are resolved relative to the preset file's directory. Use the returned
- * refs with {@link loadExtensionsIntoHost}.
+ * Paths are resolved relative to the declaring file's directory. Use the returned
+ * refs with {@link loadExtensionsIntoHost}. Rule files and presets are treated
+ * identically — both flow through the same trust gate at load time.
  */
-export function collectPresetExtensions(
-    presetName: string,
-    presetDir: string,
+export function collectExtensions(
+    sourceName: string,
+    sourceDir: string,
     extensions: Partial<Record<ExtensionKind, string[] | undefined>> | undefined,
 ): ExtensionRef[] {
     if (extensions === undefined) return [];
     const refs: ExtensionRef[] = [];
     for (const kind of ['resolvers', 'evaluators', 'fixers', 'formatters'] as ExtensionKind[]) {
         for (const path of extensions[kind] ?? []) {
-            refs.push({ kind, presetName, absPath: resolve(presetDir, path) });
+            refs.push({ kind, presetName: sourceName, absPath: resolve(sourceDir, path) });
         }
     }
     return refs;
@@ -79,8 +82,9 @@ export async function loadExtensionsIntoHost(
         );
     }
 
+    const loadModule = options.moduleLoader ?? defaultModuleLoader;
     for (const ref of refs) {
-        const moduleExports = (await import(ref.absPath)) as Record<string, unknown>;
+        const moduleExports = await loadModule(ref.absPath);
         const candidate = moduleExports.default ?? moduleExports.extension;
         if (
             candidate === null ||
@@ -106,3 +110,7 @@ export async function loadExtensionsIntoHost(
         registry.register(name, candidate, 'extension');
     }
 }
+
+async function defaultModuleLoader(absPath: string): Promise<Record<string, unknown>> {
+    return (await import(absPath)) as Record<string, unknown>;
+}

package/src/config/loader.ts

@@ -1,14 +1,15 @@
-import { basename, dirname, extname, join, relative, resolve, sep } from 'node:path';
-import { NodeFileSystem } from '@gobing-ai/ts-runtime';
-import { parse } from 'yaml';
+import { basename, dirname, join, relative, resolve, sep } from 'node:path';
+import { loadStructuredConfig, NodeFileSystem } from '@gobing-ai/ts-runtime';
 import {
     type ConstraintRule,
     type ConstraintRuleFile,
     ConstraintRuleFileSchema,
     ConstraintRuleSchema,
+    type FixMode,
     type PresetDefinition,
     PresetDefinitionSchema,
 } from '../types';
+import { collectExtensions, type ExtensionRef } from './extensions';
 
 /** Options for loading rule presets. */
 export interface RuleLoaderOptions {
@@ -20,6 +21,25 @@ export interface RuleLoaderOptions {
      * loader stays agnostic to any project layout convention.
      */
     roots: string[];
+    /** When true, honor top-level `$schema` refs in preset and rule files. Defaults to true. */
+    validateSchema?: boolean;
+    /** Optional fetch implementation for remote HTTP(S) schema refs. */
+    fetch?: (input: string) => Promise<Response>;
+}
+
+export interface RuleFileLoadOptions {
+    /** When true, honor top-level `$schema` refs. Defaults to true. */
+    validateSchema?: boolean;
+    /** Optional fetch implementation for remote HTTP(S) schema refs. */
+    fetch?: (input: string) => Promise<Response>;
+}
+
+/** Loaded preset rules plus extension module refs declared by composed presets. */
+export interface LoadedPreset {
+    /** Normalized rules after preset disable/override handling. */
+    readonly rules: ConstraintRule[];
+    /** Extension modules declared by the preset graph, resolved to absolute paths. */
+    readonly extensions: ExtensionRef[];
 }
 
 /** Merged view of rule roots: winning file per relative path, plus categories. */
@@ -37,44 +57,110 @@ interface MergedRoots {
  * so a caller can layer project-local rules over shared/global rules and inherit
  * the rest of a preset's categories from the lower-priority roots.
  */
-export async function loadPresetRules(name: string, options: RuleLoaderOptions): Promise<ConstraintRule[]> {
+export async function loadPreset(name: string, options: RuleLoaderOptions): Promise<LoadedPreset> {
     const merged = await buildMergedRoots(options.roots.map((root) => resolve(root)));
     const presetPath = findMergedPreset(merged, name);
-    if (presetPath === null) return [];
-    const preset = PresetDefinitionSchema.parse(await readStructuredFile(presetPath)) as PresetDefinition;
+    if (presetPath === null) return { rules: [], extensions: [] };
+    const preset = PresetDefinitionSchema.parse(await readStructuredFile(presetPath, options)) as PresetDefinition;
     const rules: ConstraintRule[] = [];
+    const extensions = collectExtensions(preset.name, dirname(presetPath), preset.extensions);
     for (const entry of preset.extends) {
-        rules.push(...(await loadPresetEntry(merged, entry, new Set([name]))));
+        const loaded = await loadPresetEntry(merged, entry, new Set([name]), options);
+        rules.push(...loaded.rules);
+        extensions.push(...loaded.extensions);
     }
-    const disabled = new Set(preset.disable ?? []);
-    const normalized = rules.filter((rule) => !disabled.has(rule.id));
-    for (const rule of normalized) {
-        const override = preset.overrides?.[rule.id];
+    return { rules: applyPresetControls(preset.name, rules, preset.disable, preset.overrides), extensions };
+}
+
+/** Collapse rules sharing an id to a single entry; later definitions win (last-wins merge). */
+function dedupeById(rules: readonly ConstraintRule[]): ConstraintRule[] {
+    const byId = new Map<string, ConstraintRule>();
+    for (const rule of rules) byId.set(rule.id, rule);
+    return [...byId.values()];
+}
+
+/** Apply a preset's local disable and override controls after composing its children. */
+function applyPresetControls(
+    presetName: string,
+    rules: readonly ConstraintRule[],
+    disabledIds: readonly string[] | undefined,
+    overrides: PresetDefinition['overrides'],
+): ConstraintRule[] {
+    const disabled = new Set(disabledIds ?? []);
+    const controlled = dedupeById(rules).filter((rule) => !disabled.has(rule.id));
+    for (const rule of controlled) {
+        const override = overrides?.[rule.id];
         if (override?.fix !== undefined) {
+            assertFixModeNotPromoted(presetName, rule, override.fix.mode);
             rule.fix = { ...(rule.fix ?? { mode: 'none' }), ...override.fix };
         }
     }
-    return normalized;
+    return controlled;
 }
 
-/** Load a direct rule file from disk. */
-export async function loadRuleFile(filePath: string): Promise<ConstraintRule[]> {
-    return normalizeRuleFile(await readStructuredFile(resolve(filePath)), dirname(resolve(filePath)));
+/** Fix-mode authority ordering; an override may lower but never raise a rule's mode. */
+const FIX_MODE_AUTHORITY: Record<FixMode, number> = { none: 0, suggest: 1, auto: 2 };
+
+/** Throw when a preset override would escalate a rule's fix authority above its authored level. */
+function assertFixModeNotPromoted(presetName: string, rule: ConstraintRule, overrideMode: FixMode): void {
+    const ruleMode = rule.fix?.mode ?? 'none';
+    if (FIX_MODE_AUTHORITY[overrideMode] > FIX_MODE_AUTHORITY[ruleMode]) {
+        throw new Error(
+            `Preset "${presetName}" override for rule "${rule.id}" raises fix mode from "${ruleMode}" to "${overrideMode}"; overrides may only lower fix authority`,
+        );
+    }
 }
 
-async function loadPresetEntry(merged: MergedRoots, entry: string, seen: Set<string>): Promise<ConstraintRule[]> {
+export async function loadPresetRules(name: string, options: RuleLoaderOptions): Promise<ConstraintRule[]> {
+    return (await loadPreset(name, options)).rules;
+}
+
+/**
+ * Load a direct rule file from disk.
+ *
+ * Returns the normalized rules plus any extension modules the file declares in an
+ * `extensions` block, resolved to absolute paths. Rule-file extensions are treated
+ * exactly like preset extensions — pass the refs to {@link loadExtensionsIntoHost},
+ * which enforces the same `allowExtensions` trust gate. A single-rule file (one rule
+ * object, not a `rules:` array) cannot declare extensions and yields `extensions: []`.
+ */
+export async function loadRuleFile(filePath: string, options: RuleFileLoadOptions = {}): Promise<LoadedPreset> {
+    const resolved = resolve(filePath);
+    const raw = await readStructuredFile(resolved, options);
+    const rules = normalizeRuleFile(raw, resolved);
+    const parsed = ConstraintRuleFileSchema.safeParse(raw);
+    const extensions = parsed.success
+        ? collectExtensions(basename(resolved), dirname(resolved), parsed.data.extensions)
+        : [];
+    return { rules, extensions };
+}
+
+async function loadPresetEntry(
+    merged: MergedRoots,
+    entry: string,
+    seen: Set<string>,
+    options: Pick<RuleLoaderOptions, 'validateSchema' | 'fetch'>,
+): Promise<LoadedPreset> {
     // Sub-preset reference — recurse, erroring on a genuine cycle.
     const presetPath = findMergedPreset(merged, entry);
     if (presetPath !== null) {
         if (seen.has(entry)) {
             throw new Error(`Circular preset dependency detected: ${[...seen, entry].join(' → ')}`);
         }
-        seen.add(entry);
-        const preset = PresetDefinitionSchema.safeParse(await readStructuredFile(presetPath));
+        const nextSeen = new Set([...seen, entry]);
+        const preset = PresetDefinitionSchema.safeParse(await readStructuredFile(presetPath, options));
         if (preset.success) {
             const rules: ConstraintRule[] = [];
-            for (const child of preset.data.extends) rules.push(...(await loadPresetEntry(merged, child, seen)));
-            return rules.filter((rule) => !(preset.data.disable ?? []).includes(rule.id));
+            const extensions = collectExtensions(preset.data.name, dirname(presetPath), preset.data.extensions);
+            for (const child of preset.data.extends) {
+                const loaded = await loadPresetEntry(merged, child, nextSeen, options);
+                rules.push(...loaded.rules);
+                extensions.push(...loaded.extensions);
+            }
+            return {
+                rules: applyPresetControls(preset.data.name, rules, preset.data.disable, preset.data.overrides),
+                extensions,
+            };
         }
     }
 
@@ -82,16 +168,20 @@ async function loadPresetEntry(merged: MergedRoots, entry: string, seen: Set<str
     if (merged.categories.has(entry)) {
         const rules: ConstraintRule[] = [];
         for (const absPath of mergedFilesInCategory(merged, entry)) {
-            rules.push(...(await loadRuleFile(absPath)));
+            rules.push(...normalizeRuleFile(await readStructuredFile(absPath, options), absPath));
         }
-        return rules;
+        return { rules, extensions: [] };
     }
 
     // Sub-path reference — a single winning rule file within a category.
     const subPath = findMergedFile(merged, entry);
-    if (subPath !== null) return loadRuleFile(subPath);
+    if (subPath !== null)
+        return {
+            rules: normalizeRuleFile(await readStructuredFile(subPath, options), subPath),
+            extensions: [],
+        };
 
-    return [];
+    return { rules: [], extensions: [] };
 }
 
 /**
@@ -183,30 +273,57 @@ async function listImmediateDirs(fs: NodeFileSystem, dir: string): Promise<strin
     return dirs;
 }
 
-async function readStructuredFile(path: string): Promise<unknown> {
-    const content = await new NodeFileSystem().readFile(path);
-    return extname(path) === '.json' ? JSON.parse(content) : parse(content);
+async function readStructuredFile(
+    path: string,
+    options: Pick<RuleLoaderOptions, 'validateSchema' | 'fetch'> = {},
+): Promise<unknown> {
+    return await loadStructuredConfig(path, {
+        validateSchema: options.validateSchema,
+        fetch: options.fetch,
+    });
 }
 
-function normalizeRuleFile(raw: unknown, sourceDir: string): ConstraintRule[] {
+/** A rule as parsed by Zod: severity may be absent until normalization fills it. */
+type ParsedRule = Omit<ConstraintRule, 'severity'> & { severity?: ConstraintRule['severity'] };
+
+function normalizeRuleFile(raw: unknown, filePath: string): ConstraintRule[] {
+    const sourceDir = dirname(filePath);
     const maybeFile = ConstraintRuleFileSchema.safeParse(raw);
     if (maybeFile.success) return normalizeFileRules(maybeFile.data, sourceDir);
     const maybeRule = ConstraintRuleSchema.safeParse(raw);
-    if (maybeRule.success) return [normalizeRule(maybeRule.data, {}, sourceDir)];
-    throw new Error(`Invalid rule file: ${basename(sourceDir)}`);
+    if (maybeRule.success) return [normalizeRule(maybeRule.data, sourceDir)];
+    // Surface the schema diagnostics that best fit the input: a `rules:` array means
+    // the author intended a rule file, so report against that schema; otherwise the
+    // single-rule schema. Include field paths so the offending key is obvious.
+    const isRuleFileShape = typeof raw === 'object' && raw !== null && 'rules' in raw;
+    const issues = (isRuleFileShape ? maybeFile.error : maybeRule.error).issues;
+    throw new Error(`Invalid rule file "${basename(filePath)}": ${formatIssues(issues)}`);
 }
 
-function normalizeFileRules(file: ConstraintRuleFile, sourceDir: string): ConstraintRule[] {
+/** Render Zod issues as `path: message` fragments for actionable diagnostics. */
+function formatIssues(issues: readonly { path: PropertyKey[]; message: string }[]): string {
+    return issues
+        .map((issue) => {
+            const path = issue.path.map(String).join('.');
+            return path.length > 0 ? `${path}: ${issue.message}` : issue.message;
+        })
+        .join('; ');
+}
+
+/** A rule file as parsed by Zod: rule severities may be absent until normalization. */
+type ParsedRuleFile = Omit<ConstraintRuleFile, 'rules'> & { rules: ParsedRule[] };
+
+function normalizeFileRules(file: ParsedRuleFile, sourceDir: string): ConstraintRule[] {
     return file.rules.map((rule) =>
         normalizeRule(
             {
                 ...rule,
-                severity: rule.severity ?? file.severity ?? 'error',
+                // Rule-level severity wins, then the file-level default, then 'error'.
+                severity: rule.severity ?? file.severity,
                 include: rule.include ?? file.include,
                 // File-level excludes always apply; a rule's own excludes add to (not replace) them.
                 exclude: mergeExcludes(file.exclude, rule.exclude),
             },
-            {},
             sourceDir,
         ),
     );
@@ -218,7 +335,7 @@ function mergeExcludes(fileExclude?: string[], ruleExclude?: string[]): string[]
     return [...new Set([...(fileExclude ?? []), ...(ruleExclude ?? [])])];
 }
 
-function normalizeRule(rule: ConstraintRule, _defaults: Partial<ConstraintRule>, _sourceDir: string): ConstraintRule {
+function normalizeRule(rule: ParsedRule, _sourceDir: string): ConstraintRule {
     return {
         ...rule,
         enabled: rule.enabled ?? true,

package/src/engine.ts

@@ -39,26 +39,16 @@ export class RuleEngine {
         this.host.evaluators.register(type, evaluator, 'extension');
     }
 
-    /** Evaluate all enabled rules against a working directory. */
+    /**
+     * Evaluate all enabled rules against a working directory.
+     *
+     * Thin delegate to {@link evaluateWithFixes} with `maxFixMode = 'none'`; the fix
+     * branch in that path is short-circuited by `effectiveFixMode` so callers see only
+     * findings, never auto-generated fixes. Keeps the rule loop and error-finding
+     * semantics in one place.
+     */
     async evaluate(rules: ConstraintRule[], workdir: string): Promise<RuleEngineResult> {
-        const findings: ConstraintFinding[] = [];
-        const fixes: Fix[] = [];
-        for (const rule of rules) {
-            if (rule.enabled === false) continue;
-            try {
-                const result = await this.host.evaluators.get(rule.evaluator.type).evaluate(rule, { rule, workdir });
-                findings.push(...result.findings);
-                fixes.push(...result.fixes);
-            } catch (error) {
-                findings.push(
-                    createFinding(rule, error instanceof Error ? error.message : String(error), null, {
-                        code: `evaluator:${rule.evaluator.type}`,
-                        kind: 'error',
-                    }),
-                );
-            }
-        }
-        return { findings, fixes };
+        return this.evaluateWithFixes(rules, workdir, 'none');
     }
 
     /**

package/src/evaluators/coverage-gate-evaluator.ts

@@ -100,8 +100,8 @@ export class CoverageGateEvaluator implements RuleEvaluator {
 function parseLcov(raw: string): Map<string, FileCoverage> {
     const result = new Map<string, FileCoverage>();
     let file: string | null = null;
-    let linesFound = 0;
-    let linesHit = 0;
+    let linesFound: number | null = 0;
+    let linesHit: number | null = 0;
     for (const line of raw.split('\n')) {
         const trimmed = line.trim();
         if (trimmed.startsWith('SF:')) {
@@ -109,17 +109,27 @@ function parseLcov(raw: string): Map<string, FileCoverage> {
             linesFound = 0;
             linesHit = 0;
         } else if (trimmed.startsWith('LF:')) {
-            linesFound = Number(trimmed.slice(3));
+            linesFound = parseCount(trimmed.slice(3));
         } else if (trimmed.startsWith('LH:')) {
-            linesHit = Number(trimmed.slice(3));
+            linesHit = parseCount(trimmed.slice(3));
         } else if (trimmed === 'end_of_record' && file !== null) {
-            result.set(file, { linesFound, linesHit });
+            // Skip records with malformed counts: a non-numeric LF:/LH: would
+            // otherwise yield NaN coverage and a spurious below-threshold finding.
+            if (linesFound !== null && linesHit !== null) {
+                result.set(file, { linesFound, linesHit });
+            }
             file = null;
         }
     }
     return result;
 }
 
+/** Parse an lcov count field; return null for non-finite or negative values. */
+function parseCount(raw: string): number | null {
+    const value = Number(raw);
+    return Number.isFinite(value) && value >= 0 ? value : null;
+}
+
 /** Standard exclusions applied regardless of config (tests, generated, deps). */
 function isAlwaysExcluded(filePath: string): boolean {
     return (

package/src/evaluators/file-utils.ts

@@ -34,6 +34,74 @@ export async function readWorkdirFile(workdir: string, filePath: string, fs = ne
     return await fs.readFile(resolve(workdir, filePath));
 }
 
+/** A discovered in-scope file paired with its contents. */
+export interface ScannedFile {
+    /** Workdir-relative path. */
+    readonly file: string;
+    /** Full file contents. */
+    readonly content: string;
+}
+
+/** How `scanFiles` matches `include` / `exclude` against discovered paths. */
+export type ScanMatchMode = 'loose' | 'glob';
+
+/** Options for {@link scanFiles}. */
+export interface ScanFilesOptions {
+    /** Working directory to walk. */
+    workdir: string;
+    /** Include patterns; semantics depend on `matchMode`. Undefined/empty = all files. */
+    include?: string[];
+    /** Exclude patterns; semantics depend on `matchMode`. */
+    exclude?: string[];
+    /**
+     * Scope matching policy:
+     * - `loose` — substring/suffix fragments via {@link matchesAny} (back-compat for
+     *   evaluators that historically accepted bare fragments like `.ts` or `src/`).
+     * - `glob` — anchored `**`/`*` globs via {@link matchesGlob}.
+     *
+     * The two are NOT interchangeable: a bare `src/` matches different sets under each.
+     * Each evaluator declares the mode that preserves its existing behavior.
+     */
+    matchMode: ScanMatchMode;
+    /** Filesystem adapter. */
+    fs?: FileSystem;
+}
+
+/**
+ * Discover in-scope files and read each once — the shared scaffolding behind the
+ * line-scanning evaluators. Owns discovery, scope filtering (per `matchMode`), and
+ * reads, so each evaluator is left with only its own matcher.
+ *
+ * Scope is a parameter, not assumed one-per-rule: callers that scan under several
+ * scopes (e.g. import boundaries) pass no `include` here and apply their own globs to
+ * the returned paths.
+ */
+export async function scanFiles(options: ScanFilesOptions): Promise<ScannedFile[]> {
+    const fs = options.fs ?? new NodeFileSystem();
+    const files =
+        options.matchMode === 'loose'
+            ? await discoverFiles({ workdir: options.workdir, include: options.include, exclude: options.exclude, fs })
+            : await discoverFilesByGlob(options.workdir, options.include, options.exclude, fs);
+    const scanned: ScannedFile[] = [];
+    for (const file of files) {
+        scanned.push({ file, content: await readWorkdirFile(options.workdir, file, fs) });
+    }
+    return scanned;
+}
+
+/** Discover files then filter with anchored globs (strict mode for {@link scanFiles}). */
+async function discoverFilesByGlob(
+    workdir: string,
+    include: string[] | undefined,
+    exclude: string[] | undefined,
+    fs: FileSystem,
+): Promise<string[]> {
+    const all = await discoverFiles({ workdir, fs });
+    return all
+        .filter((file) => include === undefined || include.length === 0 || include.some((g) => matchesGlob(file, g)))
+        .filter((file) => exclude === undefined || !exclude.some((g) => matchesGlob(file, g)));
+}
+
 /** Ensure a path is workdir-relative for findings. */
 export function relativeToWorkdir(workdir: string, path: string): string {
     return relative(workdir, resolve(path));
@@ -91,3 +159,27 @@ function matchSegment(segment: string, pattern: string): boolean {
     const escaped = pattern.replace(/[.+^${}()|[\]\\]/g, '\\$&').replaceAll('*', '[^/]*');
     return new RegExp(`^${escaped}$`).test(segment);
 }
+
+/** Escape a string for safe literal use inside a `RegExp` source. */
+export function escapeRegExp(value: string): string {
+    return value.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
+}
+
+/** Return the value as a `string[]` when every item is a string, otherwise undefined. */
+export function stringArray(value: unknown): string[] | undefined {
+    return Array.isArray(value) && value.every((item) => typeof item === 'string') ? (value as string[]) : undefined;
+}
+
+/**
+ * Split a leading ripgrep/PCRE-style `(?flags)` inline group off a regex source.
+ *
+ * Returns the JS-relevant flags found in the group (filtered to `imsu`) and the
+ * remaining source with the group removed. When no leading group is present, returns
+ * empty flags and the source unchanged. Shared by evaluators that accept inline flags.
+ */
+export function parseInlineFlags(source: string): { flags: string; rest: string } {
+    const match = /^\(\?([a-z]+)\)/.exec(source);
+    if (!match) return { flags: '', rest: source };
+    const flags = [...(match[1] ?? '')].filter((flag) => 'imsu'.includes(flag)).join('');
+    return { flags, rest: source.slice(match[0].length) };
+}