@nerviq/cli 1.20.0 → 1.21.0

package/src/cursor/freshness.js

@@ -41,39 +41,39 @@ const P0_SOURCES = [
     stalenessThresholdDays: 14,
     verifiedAt: '2026-04-07',
   },
-  {
-    key: 'cursor-automations',
-    label: 'Automations Documentation',
-    url: 'https://cursor.com/docs/cloud-agent/automations',
-    stalenessThresholdDays: 14,
-    verifiedAt: '2026-04-07',
-  },
-  {
-    key: 'cursor-agent-modes',
-    label: 'Cursor Agent Modes',
-    url: 'https://docs.cursor.com/en/chat/agent',
-    stalenessThresholdDays: 14,
-    verifiedAt: '2026-04-10',
-  },
-  {
-    key: 'cursor-models-docs',
-    label: 'Cursor Models & Auto Selection',
-    url: 'https://docs.cursor.com/models',
-    stalenessThresholdDays: 14,
-    verifiedAt: '2026-04-10',
-  },
-  {
-    key: 'cursor-cli-docs',
-    label: 'Cursor CLI Usage',
-    url: 'https://docs.cursor.com/en/cli/using',
-    stalenessThresholdDays: 30,
-    verifiedAt: '2026-04-10',
-  },
-  {
-    key: 'cursor-bugbot',
-    label: 'BugBot Documentation',
-    url: 'https://cursor.com/docs/bugbot',
-    stalenessThresholdDays: 30,
+  {
+    key: 'cursor-automations',
+    label: 'Automations Documentation',
+    url: 'https://cursor.com/docs/cloud-agent/automations',
+    stalenessThresholdDays: 14,
+    verifiedAt: '2026-04-07',
+  },
+  {
+    key: 'cursor-agent-modes',
+    label: 'Cursor Agent Modes',
+    url: 'https://docs.cursor.com/en/chat/agent',
+    stalenessThresholdDays: 14,
+    verifiedAt: '2026-04-10',
+  },
+  {
+    key: 'cursor-models-docs',
+    label: 'Cursor Models & Auto Selection',
+    url: 'https://docs.cursor.com/models',
+    stalenessThresholdDays: 14,
+    verifiedAt: '2026-04-10',
+  },
+  {
+    key: 'cursor-cli-docs',
+    label: 'Cursor CLI Usage',
+    url: 'https://docs.cursor.com/en/cli/using',
+    stalenessThresholdDays: 30,
+    verifiedAt: '2026-04-10',
+  },
+  {
+    key: 'cursor-bugbot',
+    label: 'BugBot Documentation',
+    url: 'https://cursor.com/docs/bugbot',
+    stalenessThresholdDays: 30,
     verifiedAt: '2026-04-07',
   },
   {
@@ -161,38 +161,38 @@ const PROPAGATION_CHECKLIST = [
       'src/cursor/deep-review.js — update trust class detection',
     ],
   },
-  {
-    trigger: 'Design Mode feature update',
-    targets: [
-      'src/cursor/setup.js — update Design Mode guide template',
-      'src/cursor/techniques.js — update CU-L01 modern features check',
-    ],
-  },
-  {
-    trigger: 'Cursor mode behavior change (Agent/Ask/Custom, auto-run, auto-fix, tool scopes)',
-    targets: [
-      'src/cursor/techniques.js — update agent-mode and autonomy checks',
-      'src/cursor/governance.js — update mode-related permission caveats',
-      'src/source-urls.js — refresh Cursor mode source mappings',
-    ],
-  },
-  {
-    trigger: 'Cursor model catalog or auto-selection change',
-    targets: [
-      'src/cursor/techniques.js — update model-awareness and cost/trust assumptions',
-      'src/cursor/governance.js — update privacy and model-selection caveats',
-      'src/source-urls.js — refresh Cursor model source mappings',
-    ],
-  },
-  {
-    trigger: 'Cursor CLI behavior change',
-    targets: [
-      'src/cursor/setup.js — update CLI-oriented starter guidance',
-      'src/cursor/techniques.js — update CLI/rules expectations where relevant',
-      'src/source-urls.js — refresh Cursor CLI source mappings',
-    ],
-  },
-];
+  {
+    trigger: 'Design Mode feature update',
+    targets: [
+      'src/cursor/setup.js — update Design Mode guide template',
+      'src/cursor/techniques.js — update CU-L01 modern features check',
+    ],
+  },
+  {
+    trigger: 'Cursor mode behavior change (Agent/Ask/Custom, auto-run, auto-fix, tool scopes)',
+    targets: [
+      'src/cursor/techniques.js — update agent-mode and autonomy checks',
+      'src/cursor/governance.js — update mode-related permission caveats',
+      'src/source-urls.js — refresh Cursor mode source mappings',
+    ],
+  },
+  {
+    trigger: 'Cursor model catalog or auto-selection change',
+    targets: [
+      'src/cursor/techniques.js — update model-awareness and cost/trust assumptions',
+      'src/cursor/governance.js — update privacy and model-selection caveats',
+      'src/source-urls.js — refresh Cursor model source mappings',
+    ],
+  },
+  {
+    trigger: 'Cursor CLI behavior change',
+    targets: [
+      'src/cursor/setup.js — update CLI-oriented starter guidance',
+      'src/cursor/techniques.js — update CLI/rules expectations where relevant',
+      'src/source-urls.js — refresh Cursor CLI source mappings',
+    ],
+  },
+];
 
 /**
  * Release gate: check if all P0 sources are within staleness threshold.

package/src/cursor/governance.js

@@ -1,229 +1,229 @@
-/**
- * Cursor Governance Module
- *
- * 6 permission profiles, 7 hook equivalents, 5 policy packs.
- *
- * Cursor-specific differences from Copilot/Gemini:
- * - 3 execution surfaces: foreground (IDE), background (cloud VM), automations (event-driven)
- * - Privacy Mode OFF by default (code sent to model providers)
- * - No terminal sandbox (unlike Copilot's VS Code sandbox)
- * - Settings stored in SQLite (state.vscdb), NOT JSON
- * - Automations run without approval — trigger misconfiguration = runaway agents
- * - MCP hard limit ~40 tools, silent drop beyond
- * - Background agent has full home directory read access (security finding Nov 2025)
- */
-
-const { CURSOR_DOMAIN_PACKS } = require('./domain-packs');
-const { CURSOR_MCP_PACKS } = require('./mcp-packs');
-
-const CURSOR_PERMISSION_PROFILES = [
-  {
-    key: 'read-only',
-    label: 'Read Only',
-    risk: 'low',
-    defaultSandbox: 'no-writes',
-    approvalPolicy: 'always-confirm',
-    useWhen: 'First contact with a repo, security review, or auditing.',
-    behavior: 'Agent can read and suggest, but all edits and terminal commands require explicit confirmation.',
-    surfaces: ['foreground'],
-  },
-  {
-    key: 'standard',
-    label: 'Standard',
-    risk: 'medium',
-    defaultSandbox: 'user-approval',
-    approvalPolicy: 'selective-approval',
-    useWhen: 'Default product work where agent edits locally but risky commands need review.',
-    behavior: 'Agent proposes edits. Terminal commands require per-command approval unless auto-run enabled.',
-    surfaces: ['foreground'],
-  },
-  {
-    key: 'agent-mode',
-    label: 'Agent Mode',
-    risk: 'medium',
-    defaultSandbox: 'auto-run-trusted',
-    approvalPolicy: 'auto-approve-safe',
-    useWhen: 'Trusted repos where full agent mode is the primary workflow.',
-    behavior: 'Full agent mode. Auto-run for trusted commands. YOLO mode available but discouraged.',
-    surfaces: ['foreground'],
-  },
-  {
-    key: 'background-agent',
-    label: 'Background Agent',
-    risk: 'high',
-    defaultSandbox: 'ephemeral-vm',
-    approvalPolicy: 'pr-gate',
-    useWhen: 'Async tasks that create branches and PRs. Privacy Mode must be disabled.',
-    behavior: 'Runs on ephemeral Ubuntu VM. Auto-approved terminal. Creates branch + PR. Full home directory readable (security finding). No max tool call limit.',
-    surfaces: ['background'],
-  },
-  {
-    key: 'automation',
-    label: 'Automation',
-    risk: 'high',
-    defaultSandbox: 'ephemeral-vm',
-    approvalPolicy: 'event-triggered',
-    useWhen: 'Event-driven triggers (push, PR, Slack, timer). No human approval before execution.',
-    behavior: 'Cloud VM per run. 30-minute cap. Diffs staged for review. Secrets via vault only. Can run hundreds per hour.',
-    surfaces: ['automations'],
-  },
-  {
-    key: 'enterprise',
-    label: 'Enterprise',
-    risk: 'low',
-    defaultSandbox: 'org-policy-enforced',
-    approvalPolicy: 'org-admin-controlled',
-    useWhen: 'Enterprise tier with SSO, SCIM, audit logs, MCP/model allowlists.',
-    behavior: 'Admin-managed policies. Privacy Mode enforced org-wide. Audit logs and AI code tracking APIs. BugBot for automated PR review.',
-    surfaces: ['foreground', 'background', 'automations'],
-  },
-];
-
-const CURSOR_HOOK_REGISTRY = [
-  {
-    key: 'always-apply-rules',
-    file: '.cursor/rules/*.mdc',
-    triggerPoint: 'alwaysApply: true',
-    matcher: 'every chat and agent message',
-    purpose: 'Inject core instructions into every agent interaction.',
-    risk: 'low',
-  },
-  {
-    key: 'auto-attached-rules',
-    file: '.cursor/rules/*.mdc',
-    triggerPoint: 'globs match',
-    matcher: 'file glob patterns',
-    purpose: 'Inject context-specific rules when matching files are open or referenced.',
-    risk: 'low',
-  },
-  {
-    key: 'auto-run-terminal',
-    file: 'Cursor Settings (SQLite)',
-    triggerPoint: 'terminal command execution',
-    matcher: 'terminal command patterns',
-    purpose: 'Auto-approve terminal commands without user confirmation (YOLO mode).',
-    risk: 'critical',
-  },
-  {
-    key: 'automation-trigger',
-    file: '.cursor/automations/*.yaml',
-    triggerPoint: 'event (push, PR, Slack, timer, webhook)',
-    matcher: 'event configuration',
-    purpose: 'Trigger cloud agent execution on specified events. Runs WITHOUT approval.',
-    risk: 'high',
-  },
-  {
-    key: 'bugbot-review',
-    file: 'BugBot config (per-repo)',
-    triggerPoint: 'PR opened/updated',
-    matcher: 'PR diff',
-    purpose: 'Automated code review on PRs. Can spawn cloud agent for auto-fixes.',
-    risk: 'medium',
-  },
-  {
-    key: 'environment-setup',
-    file: '.cursor/environment.json',
-    triggerPoint: 'background agent boot',
-    matcher: 'VM configuration',
-    purpose: 'Configure background agent VM: base image, env vars, persisted dirs, processes.',
-    risk: 'medium',
-  },
-  {
-    key: 'mcp-tool-approval',
-    file: '.cursor/mcp.json',
-    triggerPoint: 'MCP tool invocation',
-    matcher: 'tool name/server',
-    purpose: 'Control which MCP tools are available. Auto-run model for approved tools.',
-    risk: 'medium',
-  },
-];
-
-const CURSOR_POLICY_PACKS = [
-  {
-    key: 'baseline-safe',
-    label: 'Baseline Safe',
-    modules: ['.cursor/rules/ with alwaysApply', 'no .cursorrules', 'Privacy Mode awareness', 'no secrets in rules'],
-    useWhen: 'Default local Cursor rollout.',
-  },
-  {
-    key: 'background-agent-safe',
-    label: 'Background Agent Safe',
-    modules: ['environment.json configured', 'PR review gate', 'KMS secrets only', 'no direct push to main'],
-    useWhen: 'Repos using background agents for async tasks.',
-  },
-  {
-    key: 'automation-safe',
-    label: 'Automation Safe',
-    modules: ['scoped triggers (no wildcards)', 'debounce on frequent events', 'error notification', 'rate limits documented'],
-    useWhen: 'Repos using event-driven automations.',
-  },
-  {
-    key: 'enterprise-governed',
-    label: 'Enterprise Governed',
-    modules: ['SSO configured', 'audit logs enabled', 'MCP allowlist', 'Privacy Mode enforced', 'model access policy'],
-    useWhen: 'Enterprise tier repos with strict governance requirements.',
-  },
-  {
-    key: 'privacy-first',
-    label: 'Privacy First',
-    modules: ['Privacy Mode enabled', 'no secrets in any Cursor config', 'MCP env vars use ${env:} syntax', 'background agent home dir exposure documented'],
-    useWhen: 'Repos handling sensitive data where Privacy Mode and data protection are paramount.',
-  },
-];
-
-const CURSOR_PILOT_ROLLOUT_KIT = {
-  recommendedScope: [
-    'Start with audit and setup on one trusted repo before enabling background agents.',
-    'Keep .cursor/rules/ and .cursor/mcp.json in version control.',
-    'Enable Privacy Mode before using with sensitive code.',
-    'Migrate .cursorrules to .cursor/rules/*.mdc before relying on agent mode.',
-    'Test background agents on a non-critical repo first.',
-    'If using automations, start with manual triggers before event-driven.',
-  ],
-  approvals: [
-    'Engineering owner approves auto-run terminal policy and background agent usage.',
-    'Security owner approves Privacy Mode status and MCP server allowlist.',
-    'Pilot owner records before/after audit deltas and rollback expectations.',
-    'Org admin approves automation trigger scope and model access policy.',
-  ],
-  successMetrics: [
-    'Audit score delta',
-    'Surface coverage (foreground + background + automations)',
-    'Time to first useful agent task',
-    'No-overwrite rate on existing repo files',
-    'Legacy .cursorrules migration completion',
-    'MCP tool count within 40-tool limit',
-  ],
-  rollbackExpectations: [
-    'Every Cursor setup/apply write path should emit a rollback artifact.',
-    'Re-run audit after rollback to confirm the repo returned to expected state.',
-    'Background agents can be disabled by removing .cursor/environment.json.',
-    'Automations can be disabled by removing .cursor/automations/ directory.',
-  ],
-};
-
-function getCursorGovernanceSummary() {
-  return {
-    platform: 'cursor',
-    platformLabel: 'Cursor AI',
-    permissionProfiles: CURSOR_PERMISSION_PROFILES,
-    hookRegistry: CURSOR_HOOK_REGISTRY,
-    policyPacks: CURSOR_POLICY_PACKS,
-    domainPacks: CURSOR_DOMAIN_PACKS,
-    mcpPacks: CURSOR_MCP_PACKS,
-    pilotRolloutKit: CURSOR_PILOT_ROLLOUT_KIT,
-    platformCaveats: [
-      { id: 'cursorrules-ignored', severity: 'critical', message: 'Agent mode ignores .cursorrules — only .cursor/rules/*.mdc files work.' },
-      { id: 'privacy-off', severity: 'high', message: 'Privacy Mode OFF by default — code sent to model providers without zero-retention.' },
-      { id: 'mcp-tool-limit', severity: 'medium', message: 'MCP hard limit ~40 tools. Exceeding silently drops tools without warning.' },
-      { id: 'code-reversion', severity: 'critical', message: 'Code reversion risk from agent review + format-on-save + cloud sync conflict.' },
-      { id: 'no-terminal-sandbox', severity: 'high', message: 'No terminal sandbox equivalent (unlike Copilot). All terminal commands execute directly.' },
-      { id: 'bg-agent-home-dir', severity: 'critical', message: 'Background agents have full read access to home directory (~/.ssh/, ~/.aws/credentials, ~/.npmrc).' },
-      { id: 'automation-no-approval', severity: 'high', message: 'Automations run without human approval. Misconfigured triggers = uncontrolled execution.' },
-    ],
-  };
-}
-
-module.exports = {
-  getCursorGovernanceSummary,
-};
+/**
+ * Cursor Governance Module
+ *
+ * 6 permission profiles, 7 hook equivalents, 5 policy packs.
+ *
+ * Cursor-specific differences from Copilot/Gemini:
+ * - 3 execution surfaces: foreground (IDE), background (cloud VM), automations (event-driven)
+ * - Privacy Mode OFF by default (code sent to model providers)
+ * - No terminal sandbox (unlike Copilot's VS Code sandbox)
+ * - Settings stored in SQLite (state.vscdb), NOT JSON
+ * - Automations run without approval — trigger misconfiguration = runaway agents
+ * - MCP hard limit ~40 tools, silent drop beyond
+ * - Background agent has full home directory read access (security finding Nov 2025)
+ */
+
+const { CURSOR_DOMAIN_PACKS } = require('./domain-packs');
+const { CURSOR_MCP_PACKS } = require('./mcp-packs');
+
+const CURSOR_PERMISSION_PROFILES = [
+  {
+    key: 'read-only',
+    label: 'Read Only',
+    risk: 'low',
+    defaultSandbox: 'no-writes',
+    approvalPolicy: 'always-confirm',
+    useWhen: 'First contact with a repo, security review, or auditing.',
+    behavior: 'Agent can read and suggest, but all edits and terminal commands require explicit confirmation.',
+    surfaces: ['foreground'],
+  },
+  {
+    key: 'standard',
+    label: 'Standard',
+    risk: 'medium',
+    defaultSandbox: 'user-approval',
+    approvalPolicy: 'selective-approval',
+    useWhen: 'Default product work where agent edits locally but risky commands need review.',
+    behavior: 'Agent proposes edits. Terminal commands require per-command approval unless auto-run enabled.',
+    surfaces: ['foreground'],
+  },
+  {
+    key: 'agent-mode',
+    label: 'Agent Mode',
+    risk: 'medium',
+    defaultSandbox: 'auto-run-trusted',
+    approvalPolicy: 'auto-approve-safe',
+    useWhen: 'Trusted repos where full agent mode is the primary workflow.',
+    behavior: 'Full agent mode. Auto-run for trusted commands. YOLO mode available but discouraged.',
+    surfaces: ['foreground'],
+  },
+  {
+    key: 'background-agent',
+    label: 'Background Agent',
+    risk: 'high',
+    defaultSandbox: 'ephemeral-vm',
+    approvalPolicy: 'pr-gate',
+    useWhen: 'Async tasks that create branches and PRs. Privacy Mode must be disabled.',
+    behavior: 'Runs on ephemeral Ubuntu VM. Auto-approved terminal. Creates branch + PR. Full home directory readable (security finding). No max tool call limit.',
+    surfaces: ['background'],
+  },
+  {
+    key: 'automation',
+    label: 'Automation',
+    risk: 'high',
+    defaultSandbox: 'ephemeral-vm',
+    approvalPolicy: 'event-triggered',
+    useWhen: 'Event-driven triggers (push, PR, Slack, timer). No human approval before execution.',
+    behavior: 'Cloud VM per run. 30-minute cap. Diffs staged for review. Secrets via vault only. Can run hundreds per hour.',
+    surfaces: ['automations'],
+  },
+  {
+    key: 'enterprise',
+    label: 'Enterprise',
+    risk: 'low',
+    defaultSandbox: 'org-policy-enforced',
+    approvalPolicy: 'org-admin-controlled',
+    useWhen: 'Enterprise tier with SSO, SCIM, audit logs, MCP/model allowlists.',
+    behavior: 'Admin-managed policies. Privacy Mode enforced org-wide. Audit logs and AI code tracking APIs. BugBot for automated PR review.',
+    surfaces: ['foreground', 'background', 'automations'],
+  },
+];
+
+const CURSOR_HOOK_REGISTRY = [
+  {
+    key: 'always-apply-rules',
+    file: '.cursor/rules/*.mdc',
+    triggerPoint: 'alwaysApply: true',
+    matcher: 'every chat and agent message',
+    purpose: 'Inject core instructions into every agent interaction.',
+    risk: 'low',
+  },
+  {
+    key: 'auto-attached-rules',
+    file: '.cursor/rules/*.mdc',
+    triggerPoint: 'globs match',
+    matcher: 'file glob patterns',
+    purpose: 'Inject context-specific rules when matching files are open or referenced.',
+    risk: 'low',
+  },
+  {
+    key: 'auto-run-terminal',
+    file: 'Cursor Settings (SQLite)',
+    triggerPoint: 'terminal command execution',
+    matcher: 'terminal command patterns',
+    purpose: 'Auto-approve terminal commands without user confirmation (YOLO mode).',
+    risk: 'critical',
+  },
+  {
+    key: 'automation-trigger',
+    file: '.cursor/automations/*.yaml',
+    triggerPoint: 'event (push, PR, Slack, timer, webhook)',
+    matcher: 'event configuration',
+    purpose: 'Trigger cloud agent execution on specified events. Runs WITHOUT approval.',
+    risk: 'high',
+  },
+  {
+    key: 'bugbot-review',
+    file: 'BugBot config (per-repo)',
+    triggerPoint: 'PR opened/updated',
+    matcher: 'PR diff',
+    purpose: 'Automated code review on PRs. Can spawn cloud agent for auto-fixes.',
+    risk: 'medium',
+  },
+  {
+    key: 'environment-setup',
+    file: '.cursor/environment.json',
+    triggerPoint: 'background agent boot',
+    matcher: 'VM configuration',
+    purpose: 'Configure background agent VM: base image, env vars, persisted dirs, processes.',
+    risk: 'medium',
+  },
+  {
+    key: 'mcp-tool-approval',
+    file: '.cursor/mcp.json',
+    triggerPoint: 'MCP tool invocation',
+    matcher: 'tool name/server',
+    purpose: 'Control which MCP tools are available. Auto-run model for approved tools.',
+    risk: 'medium',
+  },
+];
+
+const CURSOR_POLICY_PACKS = [
+  {
+    key: 'baseline-safe',
+    label: 'Baseline Safe',
+    modules: ['.cursor/rules/ with alwaysApply', 'no .cursorrules', 'Privacy Mode awareness', 'no secrets in rules'],
+    useWhen: 'Default local Cursor rollout.',
+  },
+  {
+    key: 'background-agent-safe',
+    label: 'Background Agent Safe',
+    modules: ['environment.json configured', 'PR review gate', 'KMS secrets only', 'no direct push to main'],
+    useWhen: 'Repos using background agents for async tasks.',
+  },
+  {
+    key: 'automation-safe',
+    label: 'Automation Safe',
+    modules: ['scoped triggers (no wildcards)', 'debounce on frequent events', 'error notification', 'rate limits documented'],
+    useWhen: 'Repos using event-driven automations.',
+  },
+  {
+    key: 'enterprise-governed',
+    label: 'Enterprise Governed',
+    modules: ['SSO configured', 'audit logs enabled', 'MCP allowlist', 'Privacy Mode enforced', 'model access policy'],
+    useWhen: 'Enterprise tier repos with strict governance requirements.',
+  },
+  {
+    key: 'privacy-first',
+    label: 'Privacy First',
+    modules: ['Privacy Mode enabled', 'no secrets in any Cursor config', 'MCP env vars use ${env:} syntax', 'background agent home dir exposure documented'],
+    useWhen: 'Repos handling sensitive data where Privacy Mode and data protection are paramount.',
+  },
+];
+
+const CURSOR_PILOT_ROLLOUT_KIT = {
+  recommendedScope: [
+    'Start with audit and setup on one trusted repo before enabling background agents.',
+    'Keep .cursor/rules/ and .cursor/mcp.json in version control.',
+    'Enable Privacy Mode before using with sensitive code.',
+    'Migrate .cursorrules to .cursor/rules/*.mdc before relying on agent mode.',
+    'Test background agents on a non-critical repo first.',
+    'If using automations, start with manual triggers before event-driven.',
+  ],
+  approvals: [
+    'Engineering owner approves auto-run terminal policy and background agent usage.',
+    'Security owner approves Privacy Mode status and MCP server allowlist.',
+    'Pilot owner records before/after audit deltas and rollback expectations.',
+    'Org admin approves automation trigger scope and model access policy.',
+  ],
+  successMetrics: [
+    'Audit score delta',
+    'Surface coverage (foreground + background + automations)',
+    'Time to first useful agent task',
+    'No-overwrite rate on existing repo files',
+    'Legacy .cursorrules migration completion',
+    'MCP tool count within 40-tool limit',
+  ],
+  rollbackExpectations: [
+    'Every Cursor setup/apply write path should emit a rollback artifact.',
+    'Re-run audit after rollback to confirm the repo returned to expected state.',
+    'Background agents can be disabled by removing .cursor/environment.json.',
+    'Automations can be disabled by removing .cursor/automations/ directory.',
+  ],
+};
+
+function getCursorGovernanceSummary() {
+  return {
+    platform: 'cursor',
+    platformLabel: 'Cursor AI',
+    permissionProfiles: CURSOR_PERMISSION_PROFILES,
+    hookRegistry: CURSOR_HOOK_REGISTRY,
+    policyPacks: CURSOR_POLICY_PACKS,
+    domainPacks: CURSOR_DOMAIN_PACKS,
+    mcpPacks: CURSOR_MCP_PACKS,
+    pilotRolloutKit: CURSOR_PILOT_ROLLOUT_KIT,
+    platformCaveats: [
+      { id: 'cursorrules-ignored', severity: 'critical', message: 'Agent mode ignores .cursorrules — only .cursor/rules/*.mdc files work.' },
+      { id: 'privacy-off', severity: 'high', message: 'Privacy Mode OFF by default — code sent to model providers without zero-retention.' },
+      { id: 'mcp-tool-limit', severity: 'medium', message: 'MCP hard limit ~40 tools. Exceeding silently drops tools without warning.' },
+      { id: 'code-reversion', severity: 'critical', message: 'Code reversion risk from agent review + format-on-save + cloud sync conflict.' },
+      { id: 'no-terminal-sandbox', severity: 'high', message: 'No terminal sandbox equivalent (unlike Copilot). All terminal commands execute directly.' },
+      { id: 'bg-agent-home-dir', severity: 'critical', message: 'Background agents have full read access to home directory (~/.ssh/, ~/.aws/credentials, ~/.npmrc).' },
+      { id: 'automation-no-approval', severity: 'high', message: 'Automations run without human approval. Misconfigured triggers = uncontrolled execution.' },
+    ],
+  };
+}
+
+module.exports = {
+  getCursorGovernanceSummary,
+};