Make gx setup/init enforce a repeatable per-message agent branch lifecycle

README.md

@@ -56,6 +56,8 @@ Related tools:
 ```sh
 # inside your repo
 gx setup
+# alias:
+gx init
 ```
 
 That one command runs:
@@ -217,6 +219,7 @@ Use this exact checklist to setup multi-agent safety in this repository for Code
 
 2) Bootstrap safety in this repo:
    gx setup
+   # alias: gx init
 
    - Setup detects global OMX/OpenSpec first.
    - If one is missing and setup asks for approval, reply explicitly:
@@ -231,6 +234,9 @@ Use this exact checklist to setup multi-agent safety in this repository for Code
    bash scripts/agent-branch-start.sh "task" "agent-name"
    python3 scripts/agent-file-locks.py claim --branch "$(git rev-parse --abbrev-ref HEAD)" <file...>
    bash scripts/agent-branch-finish.sh --branch "$(git rev-parse --abbrev-ref HEAD)"
+   - For every new user message/task, repeat the same cycle:
+     start isolated agent branch/worktree -> claim file locks -> implement/verify ->
+     finish via PR/merge cleanup with scripts/agent-branch-finish.sh.
 
 5) Optional: create OpenSpec planning workspace:
    bash scripts/openspec/init-plan-workspace.sh "<plan-slug>"
@@ -251,6 +257,7 @@ Use this exact checklist to setup multi-agent safety in this repository for Code
 ```sh
 gx status [--target <path>] [--json]
 gx setup [--target <path>] [--dry-run] [--yes-global-install|--no-global-install] [--no-gitignore]
+gx init [--target <path>] [--dry-run] [--yes-global-install|--no-global-install] [--no-gitignore]
 gx doctor [--target <path>] [--dry-run] [--json] [--keep-stale-locks] [--no-gitignore]
 gx copy-prompt
 gx copy-commands
@@ -268,6 +275,7 @@ bash scripts/openspec/init-plan-workspace.sh <plan-slug>   # optional OpenSpec p
 
 No command defaults to `gx status` (non-mutating health/status view).
 `gx status` reports CLI/runtime info, global OMX/OpenSpec service status, and repo safety service state.
+`gx init` is an alias of `gx setup`.
 When run in an interactive terminal, default `GuardeX` checks npm for a newer version first
 and asks `[y/N]` whether to update immediately (default is `N`).
 

bin/multiagent-safety.js

@@ -80,6 +80,7 @@ const COMMAND_TYPO_ALIASES = new Map([
   ['realaese', 'release'],
   ['relase', 'release'],
   ['setpu', 'setup'],
+  ['inti', 'init'],
   ['intsall', 'install'],
   ['docter', 'doctor'],
   ['doctro', 'doctor'],
@@ -88,6 +89,7 @@ const COMMAND_TYPO_ALIASES = new Map([
 const SUGGESTIBLE_COMMANDS = [
   'status',
   'setup',
+  'init',
   'doctor',
   'report',
   'copy-prompt',
@@ -105,6 +107,7 @@ const SUGGESTIBLE_COMMANDS = [
 const CLI_COMMAND_DESCRIPTIONS = [
   ['status', 'Show GuardeX CLI + service health without modifying files'],
   ['setup', 'Install + repair guardrails in a git repo (supports --no-gitignore)'],
+  ['init', 'Alias of setup (bootstrap + repair guardrails in a git repo)'],
   ['doctor', 'Repair safety setup drift, then verify repo safety'],
   ['report', 'Generate security/safety reports (for example: OpenSSF scorecard)'],
   ['copy-prompt', 'Print the AI-ready setup checklist'],
@@ -127,6 +130,7 @@ const AI_SETUP_PROMPT = `Use this exact checklist to setup GuardeX (Guardian T-R
 
 2) Bootstrap safety in this repo:
    gx setup
+   # alias: gx init
 
    - Setup detects global OMX/OpenSpec first.
    - If one is missing and setup asks for approval, reply explicitly:
@@ -141,6 +145,9 @@ const AI_SETUP_PROMPT = `Use this exact checklist to setup GuardeX (Guardian T-R
    bash scripts/agent-branch-start.sh "task" "agent-name"
    python3 scripts/agent-file-locks.py claim --branch "$(git rev-parse --abbrev-ref HEAD)" <file...>
    bash scripts/agent-branch-finish.sh --branch "$(git rev-parse --abbrev-ref HEAD)"
+   - For every new user message/task, repeat the same cycle:
+     start isolated agent branch/worktree -> claim file locks -> implement/verify ->
+     finish via PR/merge cleanup with scripts/agent-branch-finish.sh.
 
 5) Optional: create OpenSpec planning workspace:
    bash scripts/openspec/init-plan-workspace.sh "<plan-slug>"
@@ -151,6 +158,9 @@ const AI_SETUP_PROMPT = `Use this exact checklist to setup GuardeX (Guardian T-R
 7) Optional: sync your current agent branch with latest base branch:
    gx sync --check
    gx sync
+
+8) Optional (GitHub remote cleanup): enable:
+   Settings -> General -> Pull Requests -> Automatically delete head branches
 `;
 
 const AI_SETUP_COMMANDS = `npm i -g @imdeadpool/guardex
@@ -272,6 +282,7 @@ ${commandCatalogLines().join('\n')}
 NOTES
   - Running ${TOOL_NAME} with no command defaults to: ${SHORT_TOOL_NAME} status
   - Short alias: ${SHORT_TOOL_NAME}
+  - ${SHORT_TOOL_NAME} init is an alias of ${SHORT_TOOL_NAME} setup
   - ${TOOL_NAME} setup asks for Y/N approval before global installs
   - Legacy command aliases are still supported: ${LEGACY_NAMES.join(', ')}`);
 
@@ -2403,7 +2414,7 @@ function main() {
     return;
   }
 
-  if (command === 'setup') {
+  if (command === 'setup' || command === 'init') {
     setup(rest);
     return;
   }

templates/AGENTS.multiagent-safety.md

@@ -10,7 +10,9 @@
 - Before deleting/replacing code, each agent must read the latest session comments/handoffs first and confirm the target code is in their owned scope.
 - If ownership is unclear or overlaps, stop that edit, post a blocker comment, and let the leader/integrator reassign scope.
 - For git isolation, each agent must start on a dedicated branch via `scripts/agent-branch-start.sh "<task-or-plan>" "<agent-name>"`.
+- Treat the base branch (`main` or the user's current local base branch) as read-only while the agent branch is active.
 - Agent completion must use `scripts/agent-branch-finish.sh` (direct merge to base when allowed; auto PR fallback for protected bases, then cleanup after merge).
+- Per-message loop is mandatory: for every new user message/task, start a fresh agent branch/worktree, claim ownership locks, implement and verify, finish via PR/merge cleanup, then repeat for the next message/task.
 
 1. Explicit ownership before edits
 

templates/codex/skills/guardex/SKILL.md

@@ -19,6 +19,8 @@ If guardrails are missing entirely, run:
 
 ```sh
 gx setup
+# alias
+gx init
 ```
 
 Then verify:
@@ -32,5 +34,6 @@ gx scan
 
 - Prefer `gx doctor` for one-step repair + verification.
 - Keep agent work isolated (`agent/*` branches + lock claims).
+- For every new user message/task, restart the full loop on a fresh agent branch/worktree.
 - For one-command Codex sandbox startup, use `bash scripts/codex-agent.sh "<task>" "<agent-name>"`.
 - Do not bypass protected branch safeguards unless explicitly required.

test/install.test.js

@@ -220,6 +220,7 @@ test('setup provisions workflow files and repo config', () => {
 
   const agentsContent = fs.readFileSync(path.join(repoDir, 'AGENTS.md'), 'utf8');
   assert.equal(agentsContent.includes('<!-- multiagent-safety:START -->'), true);
+  assert.match(agentsContent, /Per-message loop is mandatory/);
 
   const gitignoreContent = fs.readFileSync(path.join(repoDir, '.gitignore'), 'utf8');
   assert.match(gitignoreContent, /# multiagent-safety:START/);
@@ -241,6 +242,17 @@ test('setup provisions workflow files and repo config', () => {
   assert.equal(secondRun.status, 0, secondRun.stderr || secondRun.stdout);
 });
 
+test('init aliases setup and provisions workflow files', () => {
+  const repoDir = initRepo();
+
+  const result = runNode(['init', '--target', repoDir, '--no-global-install'], repoDir);
+  assert.equal(result.status, 0, result.stderr || result.stdout);
+
+  assert.equal(fs.existsSync(path.join(repoDir, 'scripts', 'agent-branch-start.sh')), true);
+  assert.equal(fs.existsSync(path.join(repoDir, 'scripts', 'agent-branch-finish.sh')), true);
+  assert.equal(fs.existsSync(path.join(repoDir, 'AGENTS.md')), true);
+});
+
 test('setup pre-commit blocks codex session commits on non-agent branches by default', () => {
   const repoDir = initRepo();
 
@@ -1187,8 +1199,10 @@ test('copy-prompt outputs AI setup instructions', () => {
   assert.match(result.stdout, /npm i -g @imdeadpool\/guardex/);
   assert.match(result.stdout, /npm i -g oh-my-codex @fission-ai\/openspec/);
   assert.match(result.stdout, /gx setup/);
+  assert.match(result.stdout, /gx init/);
   assert.match(result.stdout, /Codex or Claude/);
   assert.match(result.stdout, /scripts\/agent-file-locks.py claim/);
+  assert.match(result.stdout, /For every new user message\/task, repeat the same cycle/);
 });
 
 test('copy-commands outputs command-only checklist', () => {