feat(checks): unify inspection scope in checks/scope.ts

Single source of truth for what pygienium inspects (implementation-code extensions, exclude dirs, .d.ts/.min.* exclusions) shared by recon, dead-code, deep-modules, defensive-guards, comments, and complexity. Every scan task and agent prompt injects the shared scope rules instead of copy-pasted per-check lists.
2026-08-09 16:45:29 -04:00
parent 581436ed23
commit 2caeb2f790
10 changed files with 268 additions and 112 deletions
--- a/src/checks/comments.ts
+++ b/src/checks/comments.ts
@@ -21,6 +21,7 @@
 */

 import { registerCheck, type CheckScope } from "./registry.js";
+import { scopeRulesMarkdown } from "./scope.js";

 /** Phase-strip phase this check belongs to. */
 export const COMMENTS_PHASE_ID = "C1";
@@ -82,6 +83,7 @@ You are running the **comments** hygiene check.
 ## Target
 - Scan target: \`${scope.target}\`

+${scopeRulesMarkdown()}
 ## What to do
 1. Read the recon snapshot at \`${scope.cwd}/.pygienium/recon.json\` if it
   exists; otherwise enumerate source files directly under the target.
--- a/src/checks/complexity.ts
+++ b/src/checks/complexity.ts
@@ -22,6 +22,7 @@
 */

 import { registerCheck, type CheckScope } from "./registry.js";
+import { scopeRulesMarkdown } from "./scope.js";

 /** Phase-strip phase this check belongs to. */
 export const COMPLEXITY_PHASE_ID = "C4";
@@ -98,6 +99,7 @@ You are running the **complexity** hygiene check.
 ## Target
 - Scan target: \`${scope.target}\`

+${scopeRulesMarkdown()}
 ## What to do

 ### 1. Compute cyclomatic complexity
--- a/src/checks/dead-code.ts
+++ b/src/checks/dead-code.ts
@@ -34,6 +34,11 @@ import {
 	type CheckDefinition,
 	type CheckScope,
 } from "./registry.js";
+import {
+	isScopeSource,
+	SCOPE_EXCLUDE_DIRS,
+	scopeRulesMarkdown,
+} from "./scope.js";

 /** Dead-code categories recognised by the rubric. */
 export type DeadCategory = "export" | "file" | "shim" | "dep";
@@ -82,43 +87,6 @@ export const CATEGORY_LABELS: Record<DeadCategory, string> = {
 	dep: "Unused dependencies",
 };

-/** Source extensions worth scanning for import graphs. */
-const SOURCE_EXTENSIONS = new Set([
-	".ts",
-	".tsx",
-	".js",
-	".jsx",
-	".mjs",
-	".cjs",
-	".py",
-	".rb",
-	".go",
-	".rs",
-	".java",
-	".kt",
-	".swift",
-	".php",
-	".cs",
-	".lua",
-]);
-
-/** Directories never scanned (build output, deps, tooling). */
-const IGNORED_DIRS = new Set([
-	"node_modules",
-	".git",
-	".pygienium",
-	"dist",
-	"build",
-	"out",
-	"coverage",
-	".next",
-	".nuxt",
-	".venv",
-	"__pycache__",
-	"vendor",
-	".ralpi",
-]);
-
 /** Extensions tried when resolving a bare import specifier. */
 const RESOLVE_EXTENSIONS = [
 	".ts",
@@ -131,11 +99,6 @@ const RESOLVE_EXTENSIONS = [
 	".cts",
 ];

-const EXT = (name: string): string => {
-	const i = name.lastIndexOf(".");
-	return i === -1 ? "" : name.slice(i).toLowerCase();
-};
-
 /** Filename signals of a compat/deprecated/migration shim. */
 const SHIM_NAME_RE =
 	/(?:compat|legacy|deprecated|obsolete|obsoleted|migration|migrate|backcompat|back-compat|fallback|shim)/i;
@@ -161,7 +124,7 @@ function depStem(spec: string): string {
 async function walkSourceFiles(root: string): Promise<string[]> {
 	const st = await stat(root).catch(() => undefined);
 	if (!st) return [];
-	if (st.isFile()) return SOURCE_EXTENSIONS.has(EXT(root)) ? [root] : [];
+	if (st.isFile()) return isScopeSource(root) ? [root] : [];
 	const out: string[] = [];
 	const stack = [root];
 	while (stack.length > 0) {
@@ -175,9 +138,9 @@ async function walkSourceFiles(root: string): Promise<string[]> {
 		for (const entry of entries) {
 			const full = join(dir, entry.name);
 			if (entry.isDirectory()) {
-				if (IGNORED_DIRS.has(entry.name)) continue;
+				if (SCOPE_EXCLUDE_DIRS.has(entry.name)) continue;
 				stack.push(full);
-			} else if (entry.isFile() && SOURCE_EXTENSIONS.has(EXT(entry.name))) {
+			} else if (entry.isFile() && isScopeSource(entry.name)) {
 				out.push(full);
 			}
 		}
@@ -1069,6 +1032,8 @@ export async function buildDeadCodeScanTask(
 		"```",
 		``,
 		`Target root: ${scope.target}. Work dir: ${cwd}.`,
+		``,
+		scopeRulesMarkdown(),
 	].join("\n");
 }

--- a/src/checks/deep-modules.ts
+++ b/src/checks/deep-modules.ts
@@ -28,6 +28,7 @@ import {
 	type CheckDefinition,
 	type CheckScope,
 } from "./registry.js";
+import { isScopeSource, scopeRulesMarkdown } from "./scope.js";

 /** Output directory for this check's persistent reports. */
 export function deepModulesOutputDir(cwd: string): string {
@@ -44,26 +45,6 @@ export function changesPath(cwd: string): string {
 	return join(deepModulesOutputDir(cwd), "changes.md");
 }

-/** Source extensions this check inspects. */
-const DEEP_MODULES_EXTENSIONS = new Set([
-	".ts",
-	".tsx",
-	".js",
-	".jsx",
-	".mjs",
-	".cjs",
-	".py",
-	".rb",
-	".go",
-	".rs",
-	".java",
-	".kt",
-	".swift",
-	".php",
-	".cs",
-	".lua",
-]);
-
 /**
 * Gate: skip when the cwd has no inspectable source files at all. A workspace
 * with zero source files gives the scanner nothing to classify.
@@ -73,9 +54,7 @@ function deepModulesGate(cwd: string): string | undefined {
 	try {
 		const entries = readdirSync(cwd);
 		for (const entry of entries) {
-			const dot = entry.lastIndexOf(".");
-			const ext = dot === -1 ? "" : entry.slice(dot).toLowerCase();
-			if (DEEP_MODULES_EXTENSIONS.has(ext)) {
+			if (isScopeSource(entry)) {
 				found = true;
 				break;
 			}
@@ -116,6 +95,8 @@ function buildDeepScanTask(cwd: string, scope: CheckScope): string {
 		`count, recommendation, and risk (low if no external importers, high else).`,
 		`Then emit a one-line summary referencing the findings path.`,
 		"",
+		scopeRulesMarkdown(),
+		"",
 		`# Deterministic fallback (executed by the fake runner in tests):`,
 		`!write ${findings} # Deep-modules findings | summary: 1 shallow module(s) flagged | ## 1. ${target}/wrapper.ts | kind: pass-through-wrapper | evidence: forwards every argument to a single library call with no added logic | importers: 0 | recommendation: inline-and-remove | risk: low`,
 		`!echo deep-modules: 1 issue — see ${findings}`,
--- a/src/checks/defensive-guards.ts
+++ b/src/checks/defensive-guards.ts
@@ -48,6 +48,7 @@ import {
 	type CheckDefinition,
 	type CheckScope,
 } from "./registry.js";
+import { isScopeSource, scopeRulesMarkdown } from "./scope.js";

 /** Output directory for this check's persistent reports. */
 export function defensiveGuardsOutputDir(cwd: string): string {
@@ -64,26 +65,6 @@ export function changesPath(cwd: string): string {
 	return join(defensiveGuardsOutputDir(cwd), "changes.md");
 }

-/** Source extensions this check inspects. */
-const DEFENSIVE_GUARDS_EXTENSIONS = new Set([
-	".ts",
-	".tsx",
-	".js",
-	".jsx",
-	".mjs",
-	".cjs",
-	".py",
-	".rb",
-	".go",
-	".rs",
-	".java",
-	".kt",
-	".swift",
-	".php",
-	".cs",
-	".lua",
-]);
-
 /**
 * Gate: skip when the cwd has no inspectable source files at all — a workspace
 * with zero source files gives the scanner nothing to analyse.
@@ -93,9 +74,7 @@ function defensiveGuardsGate(cwd: string): string | undefined {
 	try {
 		const entries = readdirSync(cwd);
 		for (const entry of entries) {
-			const dot = entry.lastIndexOf(".");
-			const ext = dot === -1 ? "" : entry.slice(dot).toLowerCase();
-			if (DEFENSIVE_GUARDS_EXTENSIONS.has(ext)) {
+			if (isScopeSource(entry)) {
 				found = true;
 				break;
 			}
@@ -137,6 +116,8 @@ function buildDefensiveGuardsScanTask(cwd: string, scope: CheckScope): string {
 		`reason.`,
 		`Then emit a one-line summary referencing the findings path.`,
 		``,
+		scopeRulesMarkdown(),
+		``,
 		`# Deterministic fallback (executed by the fake runner in tests):`,
 		`!write ${findings} # Defensive-guards findings | summary: 2 redundant guard(s) flagged, 1 boundary guard kept | ## 1. ${target}/noise.ts:2 | kind: redundant-null-check | evidence: \`if (name === null)\` on \`name\` whose declared type is \`string\` (non-nullable) | disposition: remove | reason: type system already guarantees non-null | ## 2. ${target}/noise.ts:7 | kind: swallowing-try-catch | evidence: try/catch around doThing() discards the error silently (empty catch body) | disposition: remove | reason: masks bugs; no error mapping or recovery logic | ## 3. ${target}/boundary.ts:2 | kind: parsing-guard | evidence: try/catch around JSON.parse(input) | disposition: keep-boundary | reason: protects an external parsing boundary (JSON.parse of untrusted input)`,
 		`!echo defensive-guards: 2 redundant, 1 boundary kept — see ${findings}`,
--- a/src/checks/scope.ts
+++ b/src/checks/scope.ts
@@ -0,0 +1,145 @@
+/**
+ * scope.ts — canonical source-of-truth for what pygienium checks inspect.
+ *
+ * Every check (recon, dead-code, deep-modules, defensive-guards, complexity,
+ * comments) and every scanner agent prompt shares these definitions so the
+ * "only inspect implementation code" rule is stated once, not copy-pasted
+ * across four files that drift apart.
+ *
+ * @module pygienium/checks/scope
+ */
+
+/**
+ * Implementation-code file extensions pygienium inspects.
+ *
+ * Deliberately excludes documentation (`.md`, `.txt`, `.rst`), config
+ * (`.json`, `.yaml`, `.yml`, `.toml`, `.env`, `.ini`), type declarations
+ * (`.d.ts`), styles (`.css`, `.scss`), markup (`.html`, `.svg`), and lock
+ * files. These are not implementation code — a comments or complexity check
+ * flagging prose in a `.md` or a key in `package.json` is noise.
+ */
+export const SCOPE_EXTENSIONS: ReadonlySet<string> = new Set([
+	".ts",
+	".tsx",
+	".js",
+	".jsx",
+	".mjs",
+	".cjs",
+	".py",
+	".rb",
+	".go",
+	".rs",
+	".java",
+	".kt",
+	".swift",
+	".php",
+	".cs",
+	".lua",
+]);
+
+/**
+ * Directories pygienium never descends into — build output, dependency caches,
+ * tooling state, and VCS metadata. When walking the tree with `find`/`grep`/
+ * `readdir`, skip these by name to avoid wasting tokens on vendored code and
+ * generated artifacts the user can't act on.
+ */
+export const SCOPE_EXCLUDE_DIRS: ReadonlySet<string> = new Set([
+	"node_modules",
+	".git",
+	".hg",
+	".svn",
+	"dist",
+	"build",
+	"out",
+	"coverage",
+	".next",
+	".nuxt",
+	".turbo",
+	".svelte-kit",
+	"__pycache__",
+	".venv",
+	"venv",
+	"vendor",
+	".cache",
+	".pygienium",
+	".ralpi",
+	".idea",
+	".vscode",
+]);
+
+/**
+ * Compound extensions (checked after the simple extension lookup) that should
+ * be treated as non-source even though their tail extension appears in
+ * {@link SCOPE_EXTENSIONS}. The primary case: `.d.ts` type declarations are
+ * generated contracts, not implementation code.
+ */
+export const SCOPE_EXCLUDE_SUFFIXES: ReadonlySet<string> = new Set([
+	".d.ts",
+	".d.mts",
+	".d.cts",
+	".min.js",
+	".min.mjs",
+	".min.cjs",
+]);
+
+/**
+ * Test if a file path is implementation source pygienium should inspect.
+ *
+ * Returns `true` when the extension is in {@link SCOPE_EXTENSIONS} AND the
+ * path does not end with a {@link SCOPE_EXCLUDE_SUFFIXES} pattern (e.g.
+ * `.d.ts`).
+ */
+export function isScopeSource(path: string): boolean {
+	const lower = path.toLowerCase();
+	for (const suffix of SCOPE_EXCLUDE_SUFFIXES) {
+		if (lower.endsWith(suffix)) return false;
+	}
+	const dot = lower.lastIndexOf(".");
+	if (dot === -1) return false;
+	return SCOPE_EXTENSIONS.has(lower.slice(dot));
+}
+
+/**
+ * Markdown section injected into every scan task string so the sub-agent knows
+ * exactly what to inspect and what to skip — stated once here, not copy-pasted
+ * into each task builder.
+ *
+ * Agents that use `find`/`grep`/`readdir` for their own file discovery read
+ * this before exploring, so the exclusion list governs their search too.
+ */
+export function scopeRulesMarkdown(): string {
+	const extensions = [...SCOPE_EXTENSIONS]
+		.sort((a, b) => a.localeCompare(b))
+		.join("`, `");
+	const excludeDirs = [...SCOPE_EXCLUDE_DIRS]
+		.sort((a, b) => a.localeCompare(b))
+		.join("`, `");
+	return `## Scope of inspection
+
+**Only inspect implementation source files.** Do not analyse documentation,
+config, type declarations, build output, or dependencies — flagging those is
+noise the user cannot act on.
+
+### Inspect (extensions)
+\`${extensions}\`
+
+### Skip (directory names — never descend into)
+\`${excludeDirs}\`
+
+### Skip (file patterns)
+- Type declarations: \`*.d.ts\`, \`*.d.mts\`, \`*.d.cts\` — generated contracts, not impl
+- Minified bundles: \`*.min.js\`, \`*.min.mjs\`, \`*.min.cjs\` — generated, not editable
+- Docs: \`*.md\`, \`*.txt\`, \`*.rst\` — prose, not code
+- Config: \`*.json\`, \`*.yaml\`, \`*.yml\`, \`*.toml\`, \`*.ini\`, \`*.env\`
+- Styles/markup: \`*.css\`, \`*.scss\`, \`*.html\`, \`*.svg\`
+- Lock files: \`package-lock.json\`, \`*.lock\`, \`bun.lockb\`
+
+### File discovery preference
+1. **Prefer the recon snapshot** at \`<cwd>/.pygienium/recon.json\` when it
+   exists — it is the authoritative source inventory (git-tracked, extension-
+   filtered, exclude-aware). Read its \`fileCounts\` for the quick picture.
+2. Otherwise enumerate files yourself, applying the rules above.
+3. When using \`find\`/\`grep\`, add prune clauses for the skip directories
+   (e.g. \`find . -type d -name node_modules -prune -o -name '*.ts' -print\`).
+`;
+}