riskrails 3.2.1 → 4.0.1

package/.claude-plugin/marketplace.json

@@ -5,7 +5,7 @@
   },
   "metadata": {
     "description": "A weight-based engineering pipeline for Claude Code and Codex: each change is weighed on six axes, runs light, guarded or full, edits only the files its plan names, and ends at human approval — enforced by hooks, reviewed adversarially, routed to the cheap model tier by default.",
-    "version": "3.2.1"
+    "version": "4.0.1"
   },
   "plugins": [
     {
@@ -14,7 +14,7 @@
         "source": "url",
         "url": "https://github.com/vanssata/claude-agentic-sdlc-module.git"
       },
-      "version": "3.2.1",
+      "version": "4.0.1",
       "strict": true
     }
   ]

package/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "riskrails",
-  "version": "3.2.1",
+  "version": "4.0.1",
   "description": "A weight-based engineering pipeline for Claude Code and Codex: each change is weighed on six axes, runs light, guarded or full, edits only the files its plan names, and ends at human approval — enforced by hooks, reviewed adversarially, routed to the cheap model tier by default.",
   "author": {
     "name": "Ivan Kakurov"

package/.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "riskrails",
-  "version": "3.2.1",
+  "version": "4.0.1",
   "description": "A weight-based engineering pipeline for Claude Code and Codex: each change is weighed on six axes, runs light, guarded or full, edits only the files its plan names, and ends at human approval — enforced by hooks, reviewed adversarially, routed to the cheap model tier by default.",
   "author": {
     "name": "Ivan Kakurov"

package/README.md

@@ -1,6 +1,6 @@
 # RiskRails
 
-[![Version](https://img.shields.io/badge/version-3.2.1-d2f878?labelColor=101210)](https://github.com/vanssata/claude-agentic-sdlc-module/releases/tag/v3.2.1)
+[![Version](https://img.shields.io/badge/version-4.0.1-d2f878?labelColor=101210)](https://github.com/vanssata/claude-agentic-sdlc-module/releases/tag/v4.0.1)
 [![Runtimes](https://img.shields.io/badge/runtimes-Claude%20Code%20%7C%20Codex-d2f878?labelColor=101210)](docs/routing.md)
 [![Known risks](https://img.shields.io/badge/known%20risks-documented-d2f878?labelColor=101210)](docs/known-risks.md)
 [![License: MIT](https://img.shields.io/badge/license-MIT-d2f878?labelColor=101210)](LICENSE)
@@ -18,7 +18,7 @@ approval. Risky actions are stopped by a hook, not by a polite request in a
 prompt.
 
 ```bash
-npx riskrails@3.2.1   # detects Claude Code, Codex or both (Node 18+, Python 3.11+)
+npx riskrails@4.0.1   # detects Claude Code, Codex or both (Node 18+, Python 3.11+)
 ```
 
 Then, in any project: `/rr-setup` once, and `/rr-build <what you want>` for
@@ -50,83 +50,47 @@ Tuned for **one developer who knows the codebase**, on a Claude Pro, Max or
 Team plan, or a ChatGPT Plus or Pro plan, following the
 [AI-native SDLC playbook](https://claude.com/blog/the-ai-native-sdlc-playbook).
 
-**Version 3.2.1.** The guards have limits: read [Known risks](docs/known-risks.md)
+**Version 4.0.1.** The guards have limits: read [Known risks](docs/known-risks.md)
 before relying on them. What is decided but not built is in [ROADMAP.md](ROADMAP.md).
 
-## Upgrade to 3.1
-
-From 3.0.x: update the home (`npx riskrails@3.2.1`, restart the runtime), then
-in each project `/rr-setup --update` (dry run first, then `--apply`). Schema 10
-becomes **11**: migration `0011` proposes deleting `.ai/policies/risk-tiers.json`
-and `risk-tiers.md`, which `weight.json` and `weight.md` replaced; confirm with
-`--apply --confirm-delete <name>` in a terminal of your own. The update now
-renders `weight.md` from `weight.json` instead of merging it, reads the
-`docs/sdlc/*/TEMPLATE.md` files from the plugin (an unedited project copy is
-offered for deletion), and proposes cleanup of old run logs and stale tasks.
-See [CHANGELOG.md](CHANGELOG.md).
-
-## Upgrade to 3.0
-
-3.0 renames the plugin from `claude-agentic` to `riskrails`, replaces the risk
-tiers with the weight and the three modes, replaces the `/ai-*` and `/sdlc-*`
-commands with `/rr-*`, ports the guards to one Python process, and moves the
-task state to `.ai/state/state.json`. The old names stay as aliases for this
-version. Step by step, in this order:
-
-1. **Commit every project's `.ai/`** before anything else; the update tells you
-   when the tree is dirty, and `git restore` is the rollback.
-2. **Close or abandon tasks in flight.** A task started by 2.x is refused by
-   the new commands as `old surface`; finish it with the version that started
-   it, or drop it with `python3 ~/.claude/skills/ai-task/state.py close
-   --abandon`. Check with `/rr-status`.
-3. **Update the home** on every machine, for every runtime:
-
-   ```bash
-   npx riskrails@3.2.1              # or: npx -y riskrails@latest --yes
-   ```
-
-   It moves `<home>/claude-agentic/` to `<home>/riskrails/` and leaves a
-   symlink at the old path for one version, replaces the old hook entries with
-   `ai-guard.py` and `runtime-gate.py`, retires the old agents and shim files
-   to `backups/<date>/`, and rewrites the managed block. Restart the runtime.
-   In Codex, open `/hooks` and **trust the changed entries** — until you do,
-   nothing is enforced. `--dry-run` shows all of it first.
-4. **Update each project:**
-
-   ```
-   /rr-setup --update               # dry run first; read it; then --apply
-   ```
-
-   Migrations `0007` to `0010` (and `0011`, see Upgrade to 3.1) run inside the dry run, in order, before the merge:
-   `0007` creates `.ai/policies/weight.json`, carrying over the project's
-   edits to its risk-tier table; `0008` renames the managed-block markers and
-   the router's routes; `0009` moves `current.json` to `state.json`; `0010`
-   creates `.ai/policies/dialog.json` with the dialog on. Edited markdown is
-   three-way merged, JSON policies key by key with a confirmation per key;
-   `delete?` proposals are left for you to confirm. The schema becomes 10 with a 3.0 plugin, 11 with 3.1.
-5. **Decide two things the migrations cannot.** Keep the dialog (then amend
-   constitution C5 as `0010` hints) or set `"enabled": false`; and tune the
-   weight table — `domain.modules`, `customer_visible`, `hot_paths`,
-   `needs_step`, `irreversible` — then `state.py weight --mirror`.
-6. **Verify**: `/rr-status` shows schema 11 and no task in flight; a
-   `/rr-build` on a small change runs light and closes on the two facts.
-   Commit `.ai/`.
-
-Every machine of a team upgrades the plugin before a project at schema 11 is
-committed; an older plugin refuses such a project. Rollback and the full
-migration table are in [docs/project-update.md](docs/project-update.md); the
-home-side details in [docs/installation.md](docs/installation.md).
+## Upgrade to 4.0
+
+4.0 converts nothing any more: the one-version fallbacks, the migrations and
+the adopt command are gone, and `/rr-setup --update` regenerates a project's files from
+the plugin. Step by step, in this order:
+
+1. **Come from 3.1.0 or later.** 4.0 refuses a tree that still holds an old
+   shape it detects, naming the file, and regenerates any other older tree
+   without the conversions it missed. So a project last updated by an older
+   riskrails goes through the version that still converts it: install it
+   (`npx riskrails@3.1.0`), run `/rr-setup --update` with it, then install 4.0.
+2. **Commit every project's `.ai/`** first; `git restore` is the rollback.
+3. **Update the home** on every machine, for every runtime: `npx riskrails@4.0.1`
+   (the plugin route and the npm route are separate installs: reinstall both
+   if you use both). Restart the runtime. In Codex, open `/hooks` and trust the
+   changed entries.
+4. **Run `/rr-setup --update` in each project right away.** Until you do, the
+   project's old shapes are not read: the weight lacks its `domain.modules`, no
+   `verify_command` is found, and a task in the old state file is not scope-guarded.
+   Dry run first, then `--apply`; a `regenerate?` or `delete?` row is confirmed
+   by you in a terminal of your own.
+5. **Check**: `/rr-status` shows schema 12 and no task in flight.
+
+The alias commands (the `/ai-*`, `/sdlc-*`, `/project-*` and usage-report names)
+are removed; use the `/rr-*` commands. Details are in
+[docs/project-update.md](docs/project-update.md), the home-side ones in
+[docs/installation.md](docs/installation.md).
 
 ## Install
 
 ```bash
-npx riskrails@3.2.1                       # detects Claude Code, Codex, or both; proposes the detected plan
-npx riskrails@3.2.1 --target codex        # auto | claude | codex | both
-npx riskrails@3.2.1 --plan max --fable no # pro | team-pro | team-max | max | max20 | balanced-max
-npx riskrails@3.2.1 --codex-plan plus     # plus | pro | balanced-max
-npx riskrails@3.2.1 --context economy     # the plan's compaction window; the context guard also blocks
-npx riskrails@3.2.1 --statusline own      # keep your own statusline command (from 3.2.1; default: the plugin's two-line statusline)
-npx riskrails@3.2.1 --dry-run             # print what would be written, write nothing
+npx riskrails@4.0.1                       # detects Claude Code, Codex, or both; proposes the detected plan
+npx riskrails@4.0.1 --target codex        # auto | claude | codex | both
+npx riskrails@4.0.1 --plan max --fable no # pro | team-pro | team-max | max | max20 | balanced-max
+npx riskrails@4.0.1 --codex-plan plus     # plus | pro | balanced-max
+npx riskrails@4.0.1 --context economy     # the plan's compaction window; the context guard also blocks
+npx riskrails@4.0.1 --statusline own      # keep your own statusline command (from 3.2.1; default: the plugin's two-line statusline)
+npx riskrails@4.0.1 --dry-run             # print what would be written, write nothing
 npx -y riskrails@latest --yes             # update in place, keeping the recorded plan
 ```
 
@@ -145,7 +109,6 @@ The same commands in both runtimes:
 | You want to | Run |
 |---|---|
 | set up `.ai/` in a project, or bring it up to the installed plugin | `/rr-setup`, `/rr-setup --update` |
-| move another AI tool's files, or the project's own toolbox, into this layout | `/rr-setup --adopt` |
 | score the repository against the twelve plays | `/rr-setup --audit` |
 | turn an idea into intent, requirements and design | `/rr-spec <idea>` |
 | plan a spec into tasks, or one task into a mode and a start | `/rr-plan <spec or task>` |
@@ -155,10 +118,6 @@ The same commands in both runtimes:
 | close the task, get the human's approval | `/rr-ship` |
 | see where the task stands, or what it cost | `/rr-status`, `/rr-status --cost` |
 
-The old names — `/ai-init`, `/ai-task`, `/ai-status`, `/ai-audit`,
-`/project-init`, `/project-update`, `/sdlc-intent`, `/sdlc-spec`,
-`/sdlc-plan`, `/usage-report` — are aliases for one version.
-
 ### Three modes
 
 `state.py weight` scores six axes 0–2 — scope, behaviour, reversibility,
@@ -265,7 +224,6 @@ to do: [docs/known-risks.md](docs/known-risks.md).
 | Codex facts are observed, not guaranteed | grep the journals for `missed_reroute` after a Codex upgrade |
 | The dialog switch is read in two places | when the dialog and the terminal disagree, use the terminal |
 | Budgets and quota are advice, not limits | read the quota with its `stale` field |
-| `--adopt` edge cases fail safe, not smoothly | adopt on a committed tree, keep `original/` |
 | The usage report's parse cache trusts mtime and size | `--no-cache` when a number surprises you |
 
 ## Tests
@@ -276,15 +234,14 @@ python3 -I -m unittest discover -s tests -p 'test_*.py'      # CI's serial oracl
 AI_E2E=1 python3 -I -m unittest discover -s tests -p test_end_to_end.py
 ```
 
-Fifty-nine suites, 1045 tests, about 75 seconds in parallel. The guards against JSON fixtures for both runtimes and a
+Fifty-eight suites, 1133 tests, about 75 seconds in parallel. The guards against JSON fixtures for both runtimes and a
 golden of 1 341 recorded decisions, byte for byte, with timed cases that hold
 the hot path under its budget; the state machine, the weight, the batch loop,
 the dialog contract and the documents rule; the compact record and its
 pre-push validator; the installer for every plan and `--target`, the Codex
 agent renderer and the `config.toml` merge against a fixture full of
 third-party settings; `/rr-setup --update` against projects scaffolded from
-the oldest shipped templates and edited by hand; `--adopt` against Spec Kit,
-Kiro and Cursor fixtures; and an end-to-end run that installs both runtimes
+the oldest shipped templates and edited by hand; and an end-to-end run that installs both runtimes
 into scratch directories, drives one task through the guards from Claude and
 then from Codex, and proves both see the same state. No suite reads or writes
 the developer's real `~/.claude` or `~/.codex`.
@@ -295,7 +252,7 @@ the developer's real `~/.claude` or `~/.codex`.
 |---|---|
 | [docs/getting-started.md](docs/getting-started.md) | the first hour: install, `/rr-setup`, a light and a full task, a batch build, the update, the bill |
 | [docs/installation.md](docs/installation.md) | every installer option, what it installs per runtime, what loads on every turn, the limits it sets, the 2.x home-side upgrade |
-| [docs/project-update.md](docs/project-update.md) | shared versus project-owned files, how a file is updated, the migrations and their rollback, `--adopt` |
+| [docs/project-update.md](docs/project-update.md) | shared versus project-owned files, how a file is regenerated, what is kept, the rollback |
 | [docs/architecture.md](docs/architecture.md) | how the pieces fit: the installed home, `.ai/`, the pipeline, the state, the hooks, the two adapters |
 | [docs/weight.md](docs/weight.md) | the six axes, the three modes, what is measured and what is declared, tuning the table |
 | [docs/workflows.md](docs/workflows.md) | feature, bugfix, refactoring, hotfix, investigation |

package/agents/ai-reviewer.md.tmpl

@@ -54,7 +54,7 @@ in light and guarded, about 80 in full, then FINDINGS and VERDICT, with
 
 ## Boundary
 
-The brief names the task's mode; `.ai/state/state.json` (or the legacy `current.json`) has it otherwise. In
+The brief names the task's mode; `.ai/state/state.json` has it otherwise. In
 light and guarded mode your inputs are the task goal, the diff, the tests the
 step names, and the project's own fixture corpus and threat model:
 `.ai/policies/safety.md`, the guards' `*-defaults.json`, the test fixtures.

package/ai/using-riskrails.md

@@ -8,7 +8,6 @@ the steps; `policies/` holds the rules; this says which to open.
 | You want to | Run |
 |---|---|
 | set up `.ai/` in a project, or bring it up to the installed plugin | `/rr-setup`, `/rr-setup --update` |
-| move another AI tool's files into this layout | `/rr-setup --adopt` |
 | score the repository against the twelve plays | `/rr-setup --audit` |
 | turn an idea into intent, requirements and design | `/rr-spec <idea>` |
 | plan a spec into tasks, or one task into a mode and a start | `/rr-plan <spec or task>` |
@@ -18,8 +17,6 @@ the steps; `policies/` holds the rules; this says which to open.
 | close the task, get the human's approval | `/rr-ship` |
 | see where the task stands, or what it cost | `/rr-status`, `/rr-status --cost` |
 
-The old names (`/ai-task`, `/ai-init`, `/sdlc-spec` …) stay one version as aliases.
-
 ## Three modes
 
 `state.py weight` scores six axes 0–2 — scope, behaviour, reversibility, domain,

package/bin/riskrails.js

@@ -152,7 +152,6 @@ const exists = (p) => fs.existsSync(p);                       // follows symlink
 const isDir = (p) => { try { return fs.statSync(p).isDirectory(); } catch (e) { return false; } };
 const isFile = (p) => { try { return fs.statSync(p).isFile(); } catch (e) { return false; } };
 const isLink = (p) => { try { return fs.lstatSync(p).isSymbolicLink(); } catch (e) { return false; } };
-const lexists = (p) => { try { fs.lstatSync(p); return true; } catch (e) { return false; } };
 const read = (p) => fs.readFileSync(p, 'utf8');
 const readJSON = (p) => JSON.parse(read(p));
 const pretty = (v) => JSON.stringify(v, null, 2);
@@ -311,7 +310,7 @@ function renderStub(scope, runtime, dst, vars, kind = 'block') {
 // block_size — what the always-loaded block costs, and how much of the file is the user's own.
 function blockSize(file, budget) {
   const text = read(file);
-  const m = text.match(/<!-- (riskrails|claude-agentic|[Cc]odex-agentic):start -->[\s\S]*?<!-- \1:end -->/);
+  const m = text.match(/<!-- riskrails:start -->[\s\S]*?<!-- riskrails:end -->/);
   const block = m ? Buffer.byteLength(m[0]) : 0;
   out(`${path.basename(file)}: managed block ${spaced(block)} B (budget ${spaced(budget)}), file ${spaced(Buffer.byteLength(text))} B — the rest is yours`);
 }
@@ -358,12 +357,10 @@ function writeProfile(home, profile) {
 }
 
 function prevPlan(home) {
-  for (const name of ['riskrails', 'claude-agentic']) {   // before 3.4b it lived under claude-agentic/ (R35)
-    try {
-      const plan = readJSON(path.join(home, name, 'profile.json')).plan;
-      return plan === null || plan === undefined || plan === false ? '' : String(plan);
-    } catch (e) { /* next */ }
-  }
+  try {
+    const plan = readJSON(path.join(home, 'riskrails', 'profile.json')).plan;
+    return plan === null || plan === undefined || plan === false ? '' : String(plan);
+  } catch (e) { /* none */ }
   return '';
 }
 
@@ -748,15 +745,13 @@ function claudeDryRun() {
   if (C.TIER === 'max') out(`   bin:     claude-1m (opus[1m]${FABLE === 'yes' ? ' / fable[1m]' : ''}, pins the model at launch; compacts near ${C.ONE_M_AT})`);
   const plugin = claudePlugin(CLAUDE_DIR);
   skillsLine(plugin);
-  out('   retire:  skills/status (now skills/rr-status), into backups/<date>/ if present');
+  out('   retire:  skills/{status,ai-status,sdlc-intent,sdlc-plan,sdlc-spec}, into backups/<date>/ if present');
   shimLine('settings.json');
   pluginLine(plugin, true);
-  homeLine(CLAUDE_DIR);
   sharedLine();
   out('   config:  ai-git-guard.json (only if absent)');
   out('   routing: riskrails/routing.md');
   out('   profile: riskrails/profile.json');
-  out('== would migrate: the claude-routing managed block, if present, into this one');
 }
 
 // Every backup goes to one dated directory, <root>/backups/<date>/<path>: a backup beside
@@ -780,18 +775,6 @@ function retire(root, rel) {                       // move what nothing installs
   out(`retired: ${rel} -> backups/${BACKUP_DATE}/${rel}`);
 }
 
-// move_home — R35: the install home was <root>/claude-agentic/ before 3.4b. It moves to
-// <root>/riskrails/ and the old path becomes a symlink to it for one version. A real old
-// home beside an existing new one (or a link there, dangling or not) is retired.
-function moveHome(root) {
-  const old = path.join(root, 'claude-agentic'), nw = path.join(root, 'riskrails');
-  if (!isDir(old) || isLink(old)) return;
-  if (lexists(nw)) retire(root, 'claude-agentic');
-  else { fs.renameSync(old, nw); out('moved: claude-agentic/ -> riskrails/'); }
-  fs.symlinkSync('riskrails', old);
-  out('linked: claude-agentic -> riskrails (for one version)');
-}
-
 // sweep_bak — every *.bak an older installer left beside a live file moves into
 // backups/<date>/, keeping its path. Only the directories and root files this installer writes.
 function sweepBak(root, ...names) {
@@ -808,6 +791,8 @@ function sweepBak(root, ...names) {
   }
 }
 
+const RETIRED_SKILLS = ['status', 'ai-status', 'sdlc-intent', 'sdlc-plan', 'sdlc-spec'];
+
 // The seven agents the 2026-09-02 install shipped, and the nine redesign 2.4 removed (with
 // the Codex -strong/-balanced twins). Moved, not deleted: one mv out of backups/ restores one.
 const STALE_AGENTS = ['code-implementer', 'qa-runner', 'qa-fixer', 'db-migrator', 'doc-lookup', 'log-inspector',
@@ -831,8 +816,8 @@ function installFile(src, dst, root = CLAUDE_DIR) {  // back up on a real change
 // their scripts (state.py, ai-path, ...) through $AI_HOME/skills/ as on the npx route.
 function installSkills(root, plugin) {
   const backups = path.join(root, 'backups', BACKUP_DATE, 'skills');
-  // /status became /rr-status (redesign 2.3); a leftover would load as a second status command.
-  retire(root, 'skills/status');
+  // /status became /rr-status (2.3) and the ai-status and sdlc-* aliases ended in 4.0: a leftover would load as a second command.
+  for (const n of RETIRED_SKILLS) retire(root, `skills/${n}`);
   for (const name of ls(path.join(SRC, 'skills')).filter((n) => isDir(path.join(SRC, 'skills', n)))) {
     const d = path.join(SRC, 'skills', name), dst = path.join(root, 'skills', name);
     // A developer who ran the scripts in place leaves __pycache__ behind; it is not copied.
@@ -879,43 +864,15 @@ function installShared(root) {
   installFile(path.join(SRC, 'scripts/resolve-profile.py'), path.join(home, 'scripts/resolve-profile.py'), root);
 }
 
-function homeLine(root) {
-  const old = path.join(root, 'claude-agentic');
-  if (isDir(old) && !isLink(old)) out('   home:    claude-agentic/ -> riskrails/, the old path a symlink for one version');
-}
 const sharedLine = () => out(`   shared:  riskrails/ai (${countFiles(path.join(SRC, 'ai'))} files), riskrails/ai-path`);
 
-// managed_block — replace the one managed block, or append it; migrates the predecessor
-// plugin's block and an unmarked legacy section on the way. Backs up a non-empty target first.
+// managed_block — replace the one managed block, or append it. Backs up a non-empty target first.
 function managedBlock(target, blockFile, label) {
   if (!exists(target)) fs.writeFileSync(target, '');
   if (fs.statSync(target).size > 0) backupFile(target, path.dirname(target));
   const block = read(blockFile).replace(/^\n+|\n+$/g, '');
   let text = read(target);
   const start = '<!-- riskrails:start -->', end = '<!-- riskrails:end -->';
-  const notes = [];
-  const routing = text.match(/<!-- claude-routing:start -->[\s\S]*?<!-- claude-routing:end -->\n?/);
-  if (routing) {
-    text = text.slice(0, routing.index) + text.slice(routing.index + routing[0].length);
-    notes.push('migrated the claude-routing block');
-  }
-  const legacy = text.match(/^# Model allocation by task and scope[\s\S]*?(?=^# |(?![\s\S]))/m);
-  if (legacy && !text.slice(0, legacy.index).includes(start)) {
-    text = text.slice(0, legacy.index) + text.slice(legacy.index + legacy[0].length);
-    notes.push("migrated an unmarked 'Model allocation' section");
-  }
-  // A legacy claude-agentic (before 3.4b) or Codex-agentic pair is the managed block under its
-  // old name: rename it, never append beside it.
-  if (!text.includes(start) && !text.includes(end)) {
-    const pair = text.match(/<!-- (claude-agentic|[Cc]odex-agentic):start -->[\s\S]*?<!-- \1:end -->/);
-    if (pair) {
-      const name = pair[1];
-      text = text.slice(0, pair.index)
-        + pair[0].replace(`<!-- ${name}:start -->`, start).replace(`<!-- ${name}:end -->`, end)
-        + text.slice(pair.index + pair[0].length);
-      notes.push(`renamed the legacy ${name} markers`);
-    }
-  }
   let nw, action;
   if (text.includes(start) && text.includes(end)) {
     const esc = (s) => s.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
@@ -928,7 +885,7 @@ function managedBlock(target, blockFile, label) {
   }
   nw = nw.replace(/\n{4,}/g, '\n\n\n');
   fs.writeFileSync(target, nw);
-  out(`${label}: ${action}${notes.length ? ' + ' + notes.join(' + ') : ''}`);
+  out(`${label}: ${action}`);
 }
 
 // The three shell guards 3.5 retired: ai-guard.py runs all of them in one process. Their
@@ -1114,7 +1071,6 @@ const loadJSONFile = (file) => JSON.parse(read(file));
 const writeJSONFile = (file, doc) => { fs.writeFileSync(file + '.tmp', pretty(doc) + '\n'); fs.renameSync(file + '.tmp', file); };
 
 function claudeApply() {
-  moveHome(CLAUDE_DIR);
   for (const d of ['agents', 'hooks/lib', 'hooks/guard', 'skills']) mkdirp(path.join(CLAUDE_DIR, d));
   sweepBak(CLAUDE_DIR, 'settings.json', 'CLAUDE.md');
 
@@ -1390,10 +1346,9 @@ function codexDryRun() {
   out(`   hooks:   ${codexHookFiles().map((f) => path.basename(f)).join(',')}`);
   const plugin = codexPlugin(CODEX_DIR);
   skillsLine(plugin);
-  out('   retire:  skills/status (now skills/rr-status), into backups/<date>/ if present');
+  out('   retire:  skills/{status,ai-status,sdlc-intent,sdlc-plan,sdlc-spec}, into backups/<date>/ if present');
   shimLine('hooks.json');
   pluginLine(plugin, true);
-  homeLine(CODEX_DIR);
   sharedLine();
   out('   config:  ai-git-guard.json (only if absent)');
   out('   routing: riskrails/routing.md');
@@ -1404,7 +1359,6 @@ function codexDryRun() {
 }
 
 function codexApply() {
-  moveHome(CODEX_DIR);
   for (const d of ['agents', 'hooks/lib', 'hooks/guard', 'skills']) mkdirp(path.join(CODEX_DIR, d));
   sweepBak(CODEX_DIR, 'hooks.json', 'config.toml', 'AGENTS.md');
 

package/hooks/ai-path-guard-defaults.json

@@ -1,5 +1,5 @@
 {
-  "_comment": "Shipped defaults for ai-path-guard. Do not edit this file; a project overrides it in .ai/policies/path-guard.json (patterns are unioned, allow wins over every other list). Patterns are POSIX extended regular expressions matched against the absolute path. deny_patterns are secrets: refused for read and write. protected_config_patterns are always on and refuse writes: the task's state and journal, task.md, the compact record (.ai/memory/, .ai/local/memory/), and the guards' own configuration, engine and state.py under .claude/ or .codex/, in a project or in the user's home, and the riskrails plugin's cache folder, which carries the same code on the plugin route. Also the installer's statusline record (.claude/riskrails/statusline.json), which keeps the statusline command it replaced. A project's dependency_instruction_patterns, from before 3.4, are read as deny_patterns. task_protected_patterns are on only while a task is in flight (.ai/state/state.json, or else current.json, exists and has not reached \"done\"): the runtime's settings, Claude's record of installed plugins and the pipeline's policies, so a run cannot change the rules it is being judged by. Reading anything but a secret is allowed.",
+  "_comment": "Shipped defaults for ai-path-guard. Do not edit this file; a project overrides it in .ai/policies/path-guard.json (patterns are unioned, allow wins over every other list). Patterns are POSIX extended regular expressions matched against the absolute path. deny_patterns are secrets: refused for read and write. protected_config_patterns are always on and refuse writes: the task's state and journal, task.md, the compact record (.ai/memory/, .ai/local/memory/), and the guards' own configuration, engine and state.py under .claude/ or .codex/, in a project or in the user's home, and the riskrails plugin's cache folder, which carries the same code on the plugin route. Also the installer's statusline record (.claude/riskrails/statusline.json), which keeps the statusline command it replaced. task_protected_patterns are on only while a task is in flight (.ai/state/state.json, or else current.json, exists and has not reached \"done\"): the runtime's settings, Claude's record of installed plugins and the pipeline's policies, so a run cannot change the rules it is being judged by. Reading anything but a secret is allowed.",
   "deny_patterns": [
     "(^|/)\\.env$",
     "(^|/)\\.env\\.(local|prod|production|staging|live)$",

package/hooks/context-guard.py

@@ -284,7 +284,7 @@ def prompt_text(message):
     text = REMINDER.sub("", content).strip()
     # only a wrapper at the very start is a slash command: a prompt may quote these tags
     command = COMMAND.search(text) if text.startswith(("<command-name>", "<command-message>")) else None
-    if command:                     # "/ai-task fix the export" is an instruction; a bare "/model" is not
+    if command:                     # "/rr-build fix the export" is an instruction; a bare "/model" is not
         name, args = command.group(1).strip(), (COMMAND_ARGS.search(text) or [None, ""])[1].strip()
         return f"{name} {args}" if args else None
     # local command output, caveats, interruptions
@@ -360,7 +360,7 @@ def ai_root(cwd):
     """The project the session is in, or None.
 
     The walk stops at the repository it starts in, and never leaves $HOME. An
-    unbounded walk would let one `/ai-init` in a home directory adopt every
+    unbounded walk would let one `/rr-setup` in a home directory adopt every
     session on the machine: a nested, unrelated checkout would have its prompts
     recorded into the ancestor's session.json and be handed the ancestor's
     handoff at startup. cwd falls back to the process's own, which is where the
@@ -390,8 +390,6 @@ def task_in_flight(root):
     if not root:
         return False
     state = os.path.join(root, ".ai", "state", "state.json")
-    if not os.path.isfile(state):
-        state = os.path.join(root, ".ai", "state", "current.json")  # R34: legacy name, one version
     try:
         with open(state, encoding="utf-8") as fh:
             return (json.load(fh) or {}).get("current_stage") != "done"

package/hooks/guard/paths.py

@@ -672,9 +672,8 @@ class PathGuard:                            # pylint: disable=too-many-instance-
     # -------------------------------------------------------------- the task
 
     def _state_file(self):
-        """R34: state.json, or the legacy current.json for one version."""
-        new = self.root + "/.ai/state/state.json"
-        return new if is_file(new) else self.root + "/.ai/state/current.json"
+        """The task's state file."""
+        return self.root + "/.ai/state/state.json"
 
     def _task_line(self):
         loaded = load_json_file(self._state_file())
@@ -720,9 +719,7 @@ class PathGuard:                            # pylint: disable=too-many-instance-
                 if doc is not None and not isinstance(doc, dict):
                     continue
                 for tag, key in (("d", "deny_patterns"), ("a", "allow_patterns"),
-                                 ("p", "protected_config_patterns"), ("k", "task_protected_patterns"),
-                                 # A project's list from before 3.4 keeps its read and write deny.
-                                 ("d", "dependency_instruction_patterns")):
+                                 ("p", "protected_config_patterns"), ("k", "task_protected_patterns")):
                     v = doc.get(key) if doc else None
                     for item in jq_iter(v if v is not None and v is not False else []):
                         if is_jq_string(item) and item != "":

package/hooks/guard/scope.py

@@ -3,7 +3,7 @@ MultiEdit, NotebookEdit and Codex's apply_patch.
 
 During the implementation stage the approved plan names the files a step may
 touch; an edit outside them is refused with SCOPE_CHANGE_REQUIRED. Inert unless
-.ai/state/state.json (or, for one version, the legacy current.json) exists and
+.ai/state/state.json exists and
 its stage is "implementation".
 
 Bash semantics kept: a plan glob is matched as `[[ $rel == $pattern ]]` — "*"
@@ -334,8 +334,6 @@ def check(ctx):
     if root is None:
         return
     state = root + "/.ai/state/state.json"
-    if not is_file(state):
-        state = root + "/.ai/state/current.json"
     if not is_file(state):
         return
     step_id, task_id, allowed, forbidden, reason = read_state(state)