@devrik-tools/claude-gates 0.4.0 → 0.7.0

package/plugins/gates/hooks/gates/audit-before-build/index.mjs

@@ -1,88 +1,110 @@
-import { existsSync } from 'node:fs';
-import { runGate, deny, TOOL_GROUPS } from '../../lib/hook-io.mjs';
-
-const GATE_ID = 'audit-before-build';
-const CONFIG_KEY = 'requireAuditBeforeBuilding';
-
-const DEFAULT_EXECUTABLE_EXTENSIONS = [
-  '.js',
-  '.mjs',
-  '.cjs',
-  '.ts',
-  '.py',
-  '.sh',
-  '.ps1',
-];
-const DEFAULT_TOOL_FOLDERS = ['scripts/', 'hooks/', 'tools/'];
-
-const NEW_TOOL_INTENT_PATTERN =
-  /\b(create|write|build|add|cre(?:a|á)|crear|escrib(?:e|í)|escribir|constru(?:ye|í)|construir|agreg(?:a|á)|agregar)\b[^.]{0,60}\b(script|verifier|checker|gate|hook|linter|tool|verificador|chequeador|gate|hook|linter|herramienta)\b/iu;
-
-const AUDIT_EVIDENCE_PATTERN =
-  /already exists|no existing tool|no plugin|audited and|justification:|no existe una herramienta|no existe la herramienta|no hay plugin|ya existe|busque? si ya existe|verifiqu[eé] que no (hay|existe)|audit[eé] herramientas|justificacion:|justificaci[oó]n:/i;
-
-const INLINE_JUSTIFICATION_PATTERN =
-  /justification:|no-reinvent|audit:|justificacion:|justificaci[oó]n:/i;
-
-function checkDelegation(toolInput) {
-  const prompt = toolInput?.prompt;
-  if (typeof prompt !== 'string') return;
-  if (!NEW_TOOL_INTENT_PATTERN.test(prompt)) return;
-  if (AUDIT_EVIDENCE_PATTERN.test(prompt)) return;
-
-  deny(
-    GATE_ID,
-    'Delegating creation of a new script/checker/gate/hook/linter/tool without evidence of a prior audit. State what you searched and why no existing tool covers this (e.g. "audited and no existing tool...").',
-  );
-}
-
-function isNewToolFile(filePath, parameters) {
-  if (!filePath) return false;
-  const inToolFolder = parameters.toolFolders.some((folder) =>
-    filePath.includes(folder),
-  );
-  if (!inToolFolder) return false;
-  return parameters.executableExtensions.some((extension) =>
-    filePath.endsWith(extension),
-  );
-}
-
-function checkWrite(toolInput, parameters) {
-  const rawPath = toolInput?.file_path ?? '';
-  const filePath = rawPath.replace(/\\/g, '/');
-  if (!isNewToolFile(filePath, parameters)) return;
-
-  // Editing an EXISTING file is not building a new tool — only creation needs the audit.
-  // A file already on disk (any tool that carries file_path) is an edit, so it is allowed.
-  // This is what makes maintaining the gates themselves possible with the gates active.
-  if (rawPath && existsSync(rawPath)) return;
-
-  const content = toolInput?.content ?? toolInput?.new_string;
-  if (typeof content !== 'string') return;
-  if (INLINE_JUSTIFICATION_PATTERN.test(content)) return;
-
-  deny(
-    GATE_ID,
-    `Creating a new executable tool at ${filePath} without a justification comment (e.g. "justification: ..."). Document why no existing tool covers this before building a new one.`,
-  );
-}
-
-runGate(
-  {
-    id: GATE_ID,
-    configKey: CONFIG_KEY,
-    enabledByDefault: false,
-    defaultParams: {
-      executableExtensions: DEFAULT_EXECUTABLE_EXTENSIONS,
-      toolFolders: DEFAULT_TOOL_FOLDERS,
-    },
-  },
-  ({ toolName, toolInput, parameters }) => {
-    if (TOOL_GROUPS.delegation.includes(toolName)) {
-      checkDelegation(toolInput);
-      return;
-    }
-    if (!TOOL_GROUPS.write.includes(toolName)) return;
-    checkWrite(toolInput, parameters);
-  },
-);
+import { existsSync } from 'node:fs';
+import { runGate, deny, TOOL_GROUPS } from '../../lib/hook-io.mjs';
+
+const GATE_ID = 'audit-before-build';
+const CONFIG_KEY = 'requireAuditBeforeBuilding';
+
+const DEFAULT_EXECUTABLE_EXTENSIONS = [
+  '.js',
+  '.mjs',
+  '.cjs',
+  '.ts',
+  '.py',
+  '.sh',
+  '.ps1',
+];
+const DEFAULT_TOOL_FOLDERS = ['scripts/', 'hooks/', 'tools/'];
+
+// A verb ("create", "escribir", ...) followed within 60 non-period characters by a tool noun
+// ("script", "gate", ...) signals intent to build a new tool. Split into two smaller patterns
+// (checked one after the other, within a bounded window) instead of one combined alternation:
+// the single-regex form crossed sonarjs's regex-complexity budget, and this reads easier too.
+const NEW_TOOL_VERB_PATTERN =
+  /\b(?:create|write|build|add|cre[aá]r?|escrib(?:e|ir|í)|constru(?:ye|ir)|agreg[aá]r?)\b/giu;
+const NEW_TOOL_NOUN_PATTERN =
+  /\b(?:script|verifier|checker|gate|hook|linter|tool|verificador|chequeador|herramienta)\b/iu;
+const NEW_TOOL_NOUN_WINDOW = 60;
+
+function hasNewToolIntent(text) {
+  NEW_TOOL_VERB_PATTERN.lastIndex = 0;
+  let match;
+  while ((match = NEW_TOOL_VERB_PATTERN.exec(text)) !== null) {
+    const afterVerb = match.index + match[0].length;
+    let window = text.slice(afterVerb, afterVerb + NEW_TOOL_NOUN_WINDOW);
+    const dotIndex = window.indexOf('.');
+    if (dotIndex !== -1) window = window.slice(0, dotIndex);
+    if (NEW_TOOL_NOUN_PATTERN.test(window)) return true;
+    if (match.index === NEW_TOOL_VERB_PATTERN.lastIndex)
+      NEW_TOOL_VERB_PATTERN.lastIndex += 1;
+  }
+  return false;
+}
+
+const AUDIT_EVIDENCE_PATTERN =
+  /already exists|no existing tool|no plugin|audited and|justification:|no existe una herramienta|no existe la herramienta|no hay plugin|ya existe|busque? si ya existe|verifiqu[eé] que no (hay|existe)|audit[eé] herramientas|justificacion:|justificaci[oó]n:/i;
+
+const INLINE_JUSTIFICATION_PATTERN =
+  /justification:|no-reinvent|audit:|justificacion:|justificaci[oó]n:/i;
+
+function checkDelegation(toolInput) {
+  const prompt = toolInput?.prompt;
+  if (typeof prompt !== 'string') return;
+  if (!hasNewToolIntent(prompt)) return;
+  if (AUDIT_EVIDENCE_PATTERN.test(prompt)) return;
+
+  deny(
+    GATE_ID,
+    'Delegating creation of a new script/checker/gate/hook/linter/tool without evidence of a prior audit. State what you searched and why no existing tool covers this (e.g. "audited and no existing tool...").',
+  );
+}
+
+function isNewToolFile(filePath, parameters) {
+  if (!filePath) return false;
+  const inToolFolder = parameters.toolFolders.some((folder) =>
+    filePath.includes(folder),
+  );
+  if (!inToolFolder) return false;
+  return parameters.executableExtensions.some((extension) =>
+    filePath.endsWith(extension),
+  );
+}
+
+function checkWrite(toolInput, parameters) {
+  const rawPath = toolInput?.file_path ?? '';
+  const filePath = rawPath.replace(/\\/g, '/');
+  if (!isNewToolFile(filePath, parameters)) return;
+
+  // Editing an EXISTING file is not building a new tool — only creation needs the audit.
+  // A file already on disk (any tool that carries file_path) is an edit, so it is allowed.
+  // This is what makes maintaining the gates themselves possible with the gates active.
+  if (rawPath && existsSync(rawPath)) return;
+
+  const content = toolInput?.content ?? toolInput?.new_string;
+  if (typeof content !== 'string') return;
+  if (INLINE_JUSTIFICATION_PATTERN.test(content)) return;
+
+  deny(
+    GATE_ID,
+    `Creating a new executable tool at ${filePath} without a justification comment (e.g. "justification: ..."). Document why no existing tool covers this before building a new one.`,
+  );
+}
+
+runGate(
+  {
+    id: GATE_ID,
+    configKey: CONFIG_KEY,
+    enabledByDefault: false,
+    defaultParams: {
+      executableExtensions: DEFAULT_EXECUTABLE_EXTENSIONS,
+      toolFolders: DEFAULT_TOOL_FOLDERS,
+    },
+  },
+  ({ toolName, toolInput, parameters }) => {
+    if (TOOL_GROUPS.delegation.includes(toolName)) {
+      checkDelegation(toolInput);
+      return;
+    }
+    if (!TOOL_GROUPS.write.includes(toolName)) return;
+    checkWrite(toolInput, parameters);
+  },
+);

package/plugins/gates/hooks/gates/autonomous-mode/index.mjs

@@ -1,50 +1,50 @@
-// autonomous-mode — when the project turns on autonomous mode, the assistant must stop
-// asking and decide. This gate denies the AskUserQuestion tool while the mode is on, so a
-// question popup cannot interrupt an unattended run. The assistant is expected to take the
-// best, aligned decision and state the reversible assumption instead of asking.
-//
-// justification: no existing tool covers this. It is the deterministic half of "run without
-// asking": a gate can block the question tool; the prose reminder (via the session-start
-// hook and this gate's message) covers the rest, which a hook cannot force.
-//
-// ── When it acts ────────────────────────────────────────────────────────────────────
-// Only when autonomousMode is enabled in .ai/config.json (project) or the global config.
-// Off by default. The user is also reminded at session start that the mode is on (see the
-// session hook), in case they forgot to turn it off.
-//
-// ── What it does NOT do ─────────────────────────────────────────────────────────────
-// It blocks the AskUserQuestion TOOL. It cannot stop the assistant from asking in plain
-// prose (no hook sees chat text) — the injected reminder pushes against that, but the only
-// deterministic lever is the question tool, and this gate pulls it.
-
-import { runGate, deny, toolInGroups } from '../../lib/hook-io.mjs';
-
-const GATE_ID = 'autonomous-mode';
-const CONFIG_KEY = 'autonomousMode';
-
-// The tool groups whose members are "asking the user something". `toolInGroups` matches the
-// native AskUserQuestion AND any MCP ask/confirm/elicit surface (mcp__*__ask_user_*), so an
-// autonomous run can't be interrupted by a question routed through an MCP server either.
-const QUESTION_GROUPS = ['question'];
-
-const DENY_MESSAGE =
-  'Autonomous mode is ON for this project: do not ask the user, decide instead. The question ' +
-  'popup was intercepted and never shown. Now, in your VISIBLE reply, leave a one-line trace ' +
-  'so the user sees what was suppressed and what you chose — e.g. "[autonomous] I was going ' +
-  'to ask <the question>; I decided <choice>, assuming <reversible assumption>." Then take ' +
-  'that best, goal-aligned decision and proceed. Only a genuinely irreversible or dangerous ' +
-  'choice (deleting data, money, production) justifies stopping — and then stop in prose ' +
-  'saying exactly what and why, do not reach for the popup again. To let questions through, ' +
-  'set "autonomousMode": false in .ai/config.json.';
-
-runGate(
-  {
-    id: GATE_ID,
-    configKey: CONFIG_KEY,
-    enabledByDefault: false,
-  },
-  ({ toolName }) => {
-    if (!toolInGroups(toolName, QUESTION_GROUPS)) return;
-    deny(GATE_ID, DENY_MESSAGE);
-  },
-);
+// autonomous-mode — when the project turns on autonomous mode, the assistant must stop
+// asking and decide. This gate denies the AskUserQuestion tool while the mode is on, so a
+// question popup cannot interrupt an unattended run. The assistant is expected to take the
+// best, aligned decision and state the reversible assumption instead of asking.
+//
+// justification: no existing tool covers this. It is the deterministic half of "run without
+// asking": a gate can block the question tool; the prose reminder (via the session-start
+// hook and this gate's message) covers the rest, which a hook cannot force.
+//
+// ── When it acts ────────────────────────────────────────────────────────────────────
+// Only when autonomousMode is enabled in .ai/config.json (project) or the global config.
+// Off by default. The user is also reminded at session start that the mode is on (see the
+// session hook), in case they forgot to turn it off.
+//
+// ── What it does NOT do ─────────────────────────────────────────────────────────────
+// It blocks the AskUserQuestion TOOL. It cannot stop the assistant from asking in plain
+// prose (no hook sees chat text) — the injected reminder pushes against that, but the only
+// deterministic lever is the question tool, and this gate pulls it.
+
+import { runGate, deny, toolInGroups } from '../../lib/hook-io.mjs';
+
+const GATE_ID = 'autonomous-mode';
+const CONFIG_KEY = 'autonomousMode';
+
+// The tool groups whose members are "asking the user something". `toolInGroups` matches the
+// native AskUserQuestion AND any MCP ask/confirm/elicit surface (mcp__*__ask_user_*), so an
+// autonomous run can't be interrupted by a question routed through an MCP server either.
+const QUESTION_GROUPS = ['question'];
+
+const DENY_MESSAGE =
+  'Autonomous mode is ON for this project: do not ask the user, decide instead. The question ' +
+  'popup was intercepted and never shown. Now, in your VISIBLE reply, leave a one-line trace ' +
+  'so the user sees what was suppressed and what you chose — e.g. "[autonomous] I was going ' +
+  'to ask <the question>; I decided <choice>, assuming <reversible assumption>." Then take ' +
+  'that best, goal-aligned decision and proceed. Only a genuinely irreversible or dangerous ' +
+  'choice (deleting data, money, production) justifies stopping — and then stop in prose ' +
+  'saying exactly what and why, do not reach for the popup again. To let questions through, ' +
+  'set "autonomousMode": false in .ai/config.json.';
+
+runGate(
+  {
+    id: GATE_ID,
+    configKey: CONFIG_KEY,
+    enabledByDefault: false,
+  },
+  ({ toolName }) => {
+    if (!toolInGroups(toolName, QUESTION_GROUPS)) return;
+    deny(GATE_ID, DENY_MESSAGE);
+  },
+);