@idevelopers/agentlock 0.1.2 → 0.1.3

package/.claude-plugin/marketplace.json

@@ -14,7 +14,7 @@
       },
       "category": "development",
       "source": "./",
-      "version": "0.1.2",
+      "version": "0.1.3",
       "homepage": "https://www.npmjs.com/package/@idevelopers/agentlock"
     }
   ]

package/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "agentlock",
-  "version": "0.1.2",
+  "version": "0.1.3",
   "description": "Local permission firewall for AI coding agents and MCP tools",
   "author": {
     "name": "iDevelopers",

package/SECURITY.md

@@ -0,0 +1,36 @@
+# Security Policy
+
+## Supported Version
+
+AgentLock is pre-1.0 commercial MVP software. Security fixes are shipped on the latest public npm release only.
+
+| Package | Supported |
+| --- | --- |
+| `@idevelopers/agentlock@latest` | Yes |
+| Older versions | No |
+
+## Reporting Vulnerabilities
+
+Report suspected vulnerabilities privately to the repository owner or iDevelopers account owner. Do not open a public issue containing exploit details, license keys, tokens, `.env` contents, cloud credentials, browser cookies, or customer data.
+
+Please include:
+
+- affected AgentLock version
+- operating system and Node version
+- MCP client, if relevant
+- minimal reproduction steps
+- expected and actual behavior
+- redacted logs only
+
+## Security Commitments
+
+- AgentLock runs locally and does not upload local files or secrets to an AgentLock server.
+- Environment scans report risky file presence and configuration surfaces, not secret values.
+- Reports, audit logs, command output tails, and tool responses are redacted before returning to the caller.
+- Guarded commands run through `child_process.spawn` with `shell:false`.
+- Guarded command environments strip common secret-like variable names before execution.
+- Paid license validation uses Polar license keys and caches only the local activation state needed for offline grace.
+
+## Non-Goals
+
+AgentLock is not a kernel sandbox, EDR agent, malware detector, or compliance certification. It reduces accidental and agent-driven misuse through explicit local policy gates. High-risk enterprise deployments should pair AgentLock with OS sandboxing, endpoint management, secret managers, CI controls, and independent security review.

package/codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "agentlock",
-  "version": "0.1.2",
+  "version": "0.1.3",
   "description": "Local permission firewall for AI coding agents and MCP tools",
   "author": "iDevelopers",
   "license": "MIT",

package/dist/index.js

@@ -5,7 +5,7 @@ import { registerTools } from "./server.js";
 async function main() {
     const server = new McpServer({
         name: "agentlock",
-        version: "0.1.2",
+        version: "0.1.3",
     });
     registerTools(server);
     const transport = new StdioServerTransport();

package/docs/CROSS_PLATFORM_VALIDATION.md

@@ -0,0 +1,43 @@
+# Cross-Platform Validation
+
+AgentLock supports Node 20+ on macOS, Linux, and Windows.
+
+## Automated Matrix
+
+GitHub Actions runs:
+
+| OS | Node |
+| --- | --- |
+| Ubuntu latest | 20, 22, 24 |
+| macOS latest | 20, 22, 24 |
+| Windows latest | 20, 22, 24 |
+
+Each matrix job runs:
+
+```sh
+npm ci
+npm run validate
+npm run smoke:public -- @idevelopers/agentlock@latest
+```
+
+## What The Matrix Covers
+
+- TypeScript build
+- unit tests
+- npm audit gate
+- dry-run package contents
+- public npm install
+- MCP server startup
+- free scan tool
+- free policy generation tool
+- unlicensed Pro lockout with live checkout guidance
+
+## Manual Release Checks
+
+Before claiming a release is enterprise-ready, also validate:
+
+- Claude Code MCP install command
+- Claude Code plugin marketplace install
+- Codex plugin metadata import
+- Cursor MCP configuration, when supported by the customer workflow
+- one live Polar purchase and refund cycle for each changed license tier

package/docs/ENTERPRISE_READINESS.md

@@ -0,0 +1,44 @@
+# Enterprise Readiness
+
+AgentLock is live and production-validated. It is not yet independently security certified.
+
+## Completed
+
+- private source repository
+- public npm runtime package
+- public installer metadata repository
+- Polar-paid Pro and Team license model
+- live paid purchase validation
+- Pro license activation validation
+- Pro policy and guarded-command validation
+- unlicensed lockout validation
+- live cancel, refund, and revoked-license validation
+- security policy
+- threat model
+- npm audit gate
+- cross-platform CI matrix for macOS, Linux, and Windows on Node 20, 22, and 24
+
+## Claims That Are Fair Today
+
+- live production product
+- production-validated commercial MVP
+- local-first AI coding-agent firewall
+- paid-license enforcement validated against Polar
+- industry-grade foundation for agent permission control
+
+## Claims To Avoid Until External Review
+
+- independently security certified
+- SOC 2 compliant
+- ISO 27001 certified
+- formally audited
+- enterprise compliance certified
+
+## Certification Path
+
+1. Run the CI matrix on every release branch.
+2. Keep `npm audit --audit-level=moderate` clean before publish.
+3. Add signed release provenance and npm provenance publishing.
+4. Commission an independent security review focused on MCP boundaries, command execution, redaction, license enforcement, and policy bypasses.
+5. Add a public vulnerability disclosure channel.
+6. For enterprise procurement, prepare SOC 2 or ISO 27001 controls around access, change management, incident response, and vendor risk.

package/docs/PRODUCTION_REPLICATION_PLAYBOOK.md

@@ -0,0 +1,121 @@
+# AgentLock Production Runbook
+
+This app follows the AgentGate production pattern:
+
+- private source repository
+- public npm runtime package
+- public installer metadata repository
+- Polar-paid Pro/Team license key model
+- sandbox payment validation
+- live paid purchase validation
+- live cancel/refund and revoked-license validation
+
+## Product Values
+
+```text
+APP_NAME=AgentLock
+APP_SLUG=agentlock
+NPM_SCOPE=@idevelopers
+NPM_PACKAGE=@idevelopers/agentlock
+BIN_NAME=agentlock
+PRIVATE_REPO=manish-1988/agentlock
+PUBLIC_INSTALLER_REPO=manish-1988/agentlock-installer
+POLAR_ORGANIZATION_ID=534c0711-1ded-497c-aa46-d1a7dfa46d6e
+POLAR_CHECKOUT_URL=https://buy.polar.sh/polar_cl_exvU3F3ZTpiDvwBXIRqrMf30T1kBTsLxyx26B4SCkUz
+FREE_PROMISE=local risk scanner and policy generator
+PRO_PROMISE=policy enforcement, guarded commands, and audit log
+TEAM_PROMISE=shared policies, signed bundles, audit export, and seats
+```
+
+## Polar Live Release
+
+Created on 2026-08-29 in production with the iDevelopers organization default currency:
+
+```text
+CHECKOUT_LINK=AgentLock public checkout
+CHECKOUT_URL=https://buy.polar.sh/polar_cl_exvU3F3ZTpiDvwBXIRqrMf30T1kBTsLxyx26B4SCkUz
+PRO_LICENSE_BENEFIT=b39941ee-77eb-4fd7-945c-69644ec3af73
+TEAM_LICENSE_BENEFIT=11bf9e5f-6eac-43a1-9a54-f5db454f3bc2
+PRO_MONTHLY_PRODUCT=5bb20d15-8b0d-4bbd-9949-f835bdb76962
+PRO_MONTHLY_PRICE=INR 1,599/month
+PRO_ANNUAL_PRODUCT=5552305e-7c6c-4eb2-b613-0dd5ebbdf77b
+PRO_ANNUAL_PRICE=INR 15,999/year
+TEAM_MONTHLY_PRODUCT=d2bf75d5-7d5c-4848-a401-073dcc12fd08
+TEAM_MONTHLY_PRICE=INR 8,499/month
+```
+
+The Polar organization rejected USD-only pricing because the organization's default presentment currency must be included in product prices. AgentLock remains positioned for a global audience; this release uses INR because it matches the active Polar organization.
+
+## Live Payment Validation
+
+Completed on 2026-08-29 against public npm version `0.1.2`:
+
+```text
+ORDER_ID=0ee016f3-d8ac-4a7c-9b47-5f399f5583c1
+ORDER_STATUS_AFTER_PURCHASE=paid
+ORDER_STATUS_AFTER_ROLLBACK=refunded
+SUBSCRIPTION_ID=38c4b08f-8208-4baa-ba09-67960066f4bb
+SUBSCRIPTION_STATUS_AFTER_PURCHASE=active
+SUBSCRIPTION_STATUS_AFTER_ROLLBACK=canceled
+REFUND_ID=3d3f90d1-1a74-41c7-9a94-a15fd201af0e
+REFUND_STATUS=succeeded
+LICENSE_KEY_ID=777a97f5-2154-4978-8080-0b6c790363f0
+LICENSE_KEY_DISPLAY=****-04F0CB
+LICENSE_STATUS_AFTER_PURCHASE=granted
+LICENSE_STATUS_AFTER_ROLLBACK=revoked
+```
+
+Validation results:
+
+- Public `@idevelopers/agentlock@latest` installed version `0.1.2`.
+- `agentlock_activate_license` returned `tier=pro` for the purchased key.
+- `agentlock_check_action` unlocked Pro policy checks with the purchased key.
+- `agentlock_run_guarded_command` ran a harmless `node` command under Pro.
+- `agentlock_audit_log` returned the local Pro audit entries.
+- A fresh unlicensed temp HOME remained blocked from Pro tools.
+- After subscription revoke and full refund, the same key returned `tier=free` in a fresh temp HOME and Pro tools were blocked.
+
+## Validation Checklist
+
+Local source:
+
+```sh
+npm install
+npm run build
+npx vitest run
+npm pack --dry-run
+```
+
+Public npm after publish:
+
+```sh
+npm view @idevelopers/agentlock name version bin dist.tarball
+npm pack @idevelopers/agentlock --dry-run
+```
+
+Sandbox payment:
+
+1. Create AgentLock sandbox product and recurring prices.
+2. Add Polar License Key benefit with Pro metadata `tier=pro`.
+3. Buy through sandbox checkout.
+4. Activate the issued key with `agentlock_activate_license`.
+5. Confirm `agentlock_check_action` and `agentlock_run_guarded_command` unlock.
+6. Refund/cancel and verify the key no longer unlocks after revocation.
+
+Live payment:
+
+1. Create AgentLock live product and checkout. Completed 2026-08-29.
+2. Publish `@idevelopers/agentlock`. Completed 2026-08-29 with version `0.1.2`.
+3. Install from a clean temp environment. Completed 2026-08-29 against public npm version `0.1.2`.
+4. Buy the lowest paid recurring tier with a real test purchase. Completed 2026-08-29.
+5. Activate the issued key locally without pasting it into chat or docs. Completed 2026-08-29.
+6. Confirm Pro behavior works from public npm. Completed 2026-08-29.
+7. Confirm a fresh unlicensed temp HOME blocks Pro tools. Completed 2026-08-29.
+8. Cancel subscription, refund order, revoke benefit, and validate the refunded key no longer unlocks Pro. Completed 2026-08-29.
+
+## Security Notes
+
+- Never commit license keys, npm tokens, Polar server tokens, card details, OTPs, or private API keys.
+- Do not print full secret values in scan reports, audit logs, tests, docs, or command output.
+- Do not ship source in the public installer repository.
+- Keep docs honest: AgentLock is a public npm runtime with private source, not an open-source repo unless the source repo is intentionally made public.

package/docs/THREAT_MODEL.md

@@ -0,0 +1,48 @@
+# AgentLock Threat Model
+
+## Scope
+
+AgentLock protects local developer workflows where AI coding agents can request file access, shell commands, network actions, or MCP tool use. The MVP runs as a local MCP server and enforces policy for AgentLock tools.
+
+## Assets
+
+- `.env` files and local application secrets
+- cloud credentials and service account files
+- npm, GitHub, SSH, database, and deployment tokens
+- source code and customer repositories
+- local audit records
+- Polar license activation state
+
+## Trust Boundaries
+
+- AI agent prompt and tool arguments are untrusted input.
+- Local policy files are trusted only after human review.
+- Polar is trusted for paid license status.
+- npm is trusted for public package distribution.
+- The local OS user account is outside the MVP enforcement boundary.
+
+## Primary Threats And Controls
+
+| Threat | Control |
+| --- | --- |
+| Agent reads secret files | default deny patterns for `.env`, cloud credentials, SSH keys, npm tokens, browser cookies |
+| Agent runs destructive shell commands | command classification, deny patterns, approval-required categories |
+| Agent publishes packages or pushes code | explicit package-publish and git-push categories |
+| Agent exfiltrates data over network | network-egress category and domain allow/deny policy |
+| Agent overuses MCP tools | MCP server and tool allowlists |
+| Agent output leaks secrets | centralized redaction for tool responses, audit entries, and command output tails |
+| License bypass after refund | live Polar validation plus revoked-license fallback to Free |
+| Token leakage through child process env | guarded command environment strips secret-like variable names |
+
+## Residual Risks
+
+- AgentLock only enforces actions routed through AgentLock tools.
+- A user can still run unsafe commands outside AgentLock.
+- Local malware or a compromised OS account can bypass the MCP boundary.
+- Policy mistakes can over-allow risky actions.
+- Offline grace allows recent valid licenses to keep working during network failures.
+- Independent audit and formal compliance certification have not yet been completed.
+
+## Validation Evidence
+
+The production runbook records the live purchase, Pro activation, public npm smoke, refund, subscription cancellation, and revoked-license validation for AgentLock `0.1.2`.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@idevelopers/agentlock",
-  "version": "0.1.2",
+  "version": "0.1.3",
   "description": "Local permission firewall for AI coding agents and MCP tools.",
   "type": "module",
   "bin": {
@@ -9,11 +9,16 @@
   "scripts": {
     "build": "tsc",
     "prepare": "npm run build",
-    "test": "vitest run"
+    "test": "vitest run",
+    "audit:security": "npm audit",
+    "smoke:public": "node scripts/public-package-smoke.mjs",
+    "validate": "npm run build && npm run test && npm run audit:security && npm pack --dry-run"
   },
   "files": [
     "dist",
     "README.md",
+    "SECURITY.md",
+    "docs",
     "agentlock.policy.example.json",
     ".mcp.json",
     ".claude-plugin",

package/skills/agentlock-workflow/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: agentlock-workflow
 description: Use when running AI coding work under AgentLock permission capsules, secret-safe scans, command checks, or guarded command execution.
-version: 0.1.2
+version: 0.1.3
 ---
 
 # AgentLock Workflow