claudemd-cli 0.9.1

package/scripts/lib/lint.js

@@ -0,0 +1,128 @@
+// lint.js — pure-Node scanning functions for §10-V banned-vocab patterns.
+// Shared by bin/claudemd-lint.js (CLI) and any future Node-side enforcement
+// that doesn't want to shell out to bash. Mirrors the substantive matching
+// rules of hooks/banned-vocab-check.sh + hooks/transcript-vocab-scan.sh
+// without the shell-specific quoting + jq plumbing.
+//
+// The patterns file is the authoritative source — hooks/banned-vocab.patterns.
+// One regex per non-blank, non-comment line. Format:
+//   <extended-regex>|<reason>
+//   <extended-regex>|@ratio <reason>     ← ratio class, exempt under baseline
+//
+// JS regex notes:
+//   * Patterns were authored for grep -iE (POSIX ERE). Most carry over to
+//     JS regex unchanged. `\b` and `[0-9]` are equivalent. POSIX char classes
+//     like [[:space:]] are NOT in the .patterns file.
+//   * `\s` means whitespace in JS — also fine.
+//   * Case-insensitive matching is the contract; we always pass /i flag.
+//   * Invalid regex (a future bad pattern checked in by mistake) is skipped
+//     silently rather than crashing the scan — fail-open consistent with
+//     the bash hooks' design.
+
+import fs from 'node:fs';
+import path from 'node:path';
+import { fileURLToPath } from 'node:url';
+
+const HERE = path.dirname(fileURLToPath(import.meta.url));
+export const DEFAULT_PATTERNS_FILE = path.resolve(HERE, '../../hooks/banned-vocab.patterns');
+
+// Returns [{regex: string, reason: string, isRatio: boolean}, ...].
+export function readPatterns(patternsFile = DEFAULT_PATTERNS_FILE) {
+  if (!fs.existsSync(patternsFile)) return [];
+  const lines = fs.readFileSync(patternsFile, 'utf8').split('\n');
+  const out = [];
+  for (const raw of lines) {
+    const line = raw.trim();
+    if (!line || line.startsWith('#')) continue;
+    // Right-most `|` is the separator: regex bodies can themselves contain `|`
+    // (alternation), but reason text doesn't. Same convention banned-vocab-
+    // check.sh uses (`${line%|*}` / `${line##*|}`).
+    const lastBar = line.lastIndexOf('|');
+    if (lastBar === -1) continue;
+    const regex = line.slice(0, lastBar);
+    let reason = line.slice(lastBar + 1);
+    let isRatio = false;
+    if (reason.startsWith('@ratio ')) {
+      isRatio = true;
+      reason = reason.slice('@ratio '.length);
+    }
+    out.push({ regex, reason, isRatio });
+  }
+  return out;
+}
+
+// scan(text, opts) → [{match, regex, reason, isRatio}, ...]
+//   opts.excludeRatio: skip @ratio-tagged patterns. The bash transcript scan
+//     does this because chat prose uses ratios with different baseline
+//     conventions than commit messages. CLI `lint` defaults to NO exclude
+//     (commit-message context is the most common use); CLI `audit` defaults
+//     to excludeRatio=true to mirror transcript-vocab-scan.sh behavior.
+//   opts.patterns: pre-loaded patterns array (lets callers cache the read).
+export function scan(text, { excludeRatio = false, patterns } = {}) {
+  if (!text) return [];
+  const pats = patterns || readPatterns();
+  const hits = [];
+  for (const p of pats) {
+    if (excludeRatio && p.isRatio) continue;
+    let re;
+    try {
+      re = new RegExp(p.regex, 'i');
+    } catch {
+      continue; // bad regex — skip (fail-open)
+    }
+    const m = text.match(re);
+    if (m) hits.push({ match: m[0], regex: p.regex, reason: p.reason, isRatio: p.isRatio });
+  }
+  return hits;
+}
+
+// parseTranscript(jsonlText) → [{turnIndex, line, text}, ...]
+//   Iterates jsonl, returns one entry per assistant text-content turn. Each
+//   entry concatenates all .message.content[*].text blocks for that turn.
+//   Corrupt rows (unparseable JSON, missing fields) silently skipped — matches
+//   transcript-vocab-scan.sh's `try fromjson catch empty` design.
+export function parseTranscript(jsonlText) {
+  const lines = jsonlText.split('\n');
+  const turns = [];
+  let turnIndex = 0;
+  for (let i = 0; i < lines.length; i++) {
+    if (!lines[i].trim()) continue;
+    let row;
+    try { row = JSON.parse(lines[i]); } catch { continue; }
+    if (row.type !== 'assistant') continue;
+    const content = row.message?.content || [];
+    const texts = [];
+    for (const b of content) {
+      if (b.type === 'text' && typeof b.text === 'string') texts.push(b.text);
+    }
+    if (texts.length === 0) continue;
+    turns.push({ turnIndex, line: i + 1, text: texts.join(' ') });
+    turnIndex++;
+  }
+  return turns;
+}
+
+// Format helpers — keep the CLI thin.
+export function formatHumanReadable({ scope, hits, turns }) {
+  if (scope === 'lint') {
+    if (hits.length === 0) return 'OK: no §10-V hits';
+    const lines = [`§10-V drift detected (${hits.length} hit${hits.length === 1 ? '' : 's'}):`];
+    for (const h of hits) lines.push(`  - "${h.match}"  (${h.reason})`);
+    return lines.join('\n');
+  }
+  if (scope === 'audit') {
+    const flagged = turns.filter(t => t.hits.length > 0);
+    if (flagged.length === 0) return `OK: no §10-V hits across ${turns.length} assistant turn(s)`;
+    const lines = [`§10-V drift detected in ${flagged.length} of ${turns.length} assistant turn(s):`];
+    for (const t of flagged) {
+      lines.push(`  line ${t.line} (turn #${t.turnIndex}):`);
+      for (const h of t.hits) lines.push(`    - "${h.match}"  (${h.reason})`);
+    }
+    return lines.join('\n');
+  }
+  return '';
+}
+
+export function formatJSON(payload) {
+  return JSON.stringify(payload, null, 2);
+}