coderifts 4.4.0 → 4.6.0

package/CHANGELOG.md

@@ -1,5 +1,71 @@
 # Changelog
 
+## 4.6.0
+
+**Version pick:** the parallel Cursor-hook round claimed **4.5.0**
+(`8eab5f9`, `packages/cli/package.json` 4.5.0). This command takes the next
+free minor. (4.5.0's CHANGELOG entry was not written in that round; the
+commit message is the record of `coderifts cursor-hook` + `hook install
+--cursor`.)
+
+### Added
+- **`coderifts init --agents`** — one command that wires a repository for
+  governed agent work. Four pieces:
+  1. repo-level MCP config per host (never one file for all):
+     Claude Code `.mcp.json` uses `mcpServers.coderifts` with **`type":"http"`
+     + `url`** (a `url` without `type` is skipped — measured
+     code.claude.com/docs/en/mcp 2026-08-22); Cursor `.cursor/mcp.json` uses
+     **`url` and no `type`** (measured docs.cursor.com). Gemini `httpUrl` is
+     extension-level — this command does not write it.
+  2. generated agent rule files, via **`writeAgentHostFiles`** extracted from
+     `agent-setup` (not a second copy). Cursor `.mdc` is included because the
+     4.5.0 round landed it.
+  3. host hook where one exists: `installClaudeHook` and (4.5.0)
+     `installCursorHook` — JSON-merge, `.bak`, idempotent. `--no-hook` skips.
+  4. `.github/workflows/coderifts.yml` at `coderifts/contract-gate@v0`, with
+     `require-verified-monitoring` commented (honest default: false).
+     `--no-workflow` skips. Existing workflow files are never clobbered.
+- `--hosts=claude,cursor,all` (default `all`), `--dry-run` (plan, no write),
+  `--check` (which of the four pieces are present).
+- Idempotent re-run: unchanged files are skipped; JSON MCP configs are merged
+  (other servers preserved, `.bak` alongside any file it modifies).
+- Summary names created / merged / skipped and WHY, the next manual step
+  (add `CODERIFTS_API_KEY` to repo secrets — the CLI never writes a key),
+  and what this does **not** do: enable branch protection
+  (ENFORCEMENT.md), make raw host tools inescapable, or enforce the
+  workflow until the required check is turned on in repo settings.
+
+### Shared vs new
+- **Shared:** `writeAgentHostFiles` (extracted from `agent-setup`);
+  `installClaudeHook` / `installCursorHook` (`silent` + `dryRun` passed
+  through).
+- **New:** `src/mcp-repo-config.js`, `src/contract-gate-workflow.js`,
+  `src/commands/init-agents.js`.
+
+## 4.4.1
+
+**Sequencing:** `@coderifts/agent-guard` **9.0.0** is the verifier of record
+for `deploy-gate` (TOKEN mode). Publish **guard 9.0.0 first**, then this CLI.
+4.4.0 against 9.0.0 would DENY `unverified_receipt_view` (unstamped computed
+view).
+
+### Changed
+- **TOKEN mode.** `deployBind` passes `{ token, decision_result, registry |
+  pinnedKeyPem }` to `deployGate`. The CLI never sends an unstamped
+  `currently_authorized` view. The §154 matrix view path is stamped with
+  `asVerifiedDeployReceiptView` (`view_spec: 'deploy-receipt-view.v1'`).
+- **One verdict, one WHY.** Guard 9.0.0 is the cause. The app kernel still
+  verifies for 842 rendering (corroboration + envelope `blocking_reasons`)
+  but does not produce a second deny. `authz_reason` equals the gate
+  `reason`. JSON `causes_agree` is true when kernel maps to the same cause.
+- **`MIN_GUARD_MAJOR` 8 → 9.** Dependency range `^9.0.0`.
+
+### Migration
+- Requires `@coderifts/agent-guard@9.0.0`. Local test against unpublished
+  9.0.0: install the packed tarball into `node_modules` while keeping the
+  declared range `^9.0.0` (`file:` is rejected by `assert-guard-major` at
+  publish).
+
 ## 4.4.0
 
 **Behaviour change (breaking for anyone relying on a forged receipt file)** —

package/README.md

@@ -61,6 +61,42 @@ Interactive configuration generator. Creates a `.coderifts.yml` file with indust
 - `microservices-internal`
 - `ai-agent-platform`
 
+### `coderifts init --agents`
+
+One command that wires **this repository** for governed agent work. Re-run is
+a no-op. Existing JSON is merged (never clobbered); a `.bak` sits next to any
+file it modifies.
+
+```bash
+coderifts init --agents                  # all hosts, all four pieces
+coderifts init --agents --dry-run        # plan only — writes nothing
+coderifts init --agents --hosts=claude   # Claude Code only
+coderifts init --agents --hosts=cursor   # Cursor only
+coderifts init --agents --no-hook        # skip host hook install
+coderifts init --agents --no-workflow    # skip .github/workflows/coderifts.yml
+coderifts init --agents --check          # which of the four pieces are present
+```
+
+| Piece | What it writes | Host-specific shape |
+|-------|----------------|---------------------|
+| **MCP config** | `.mcp.json` (Claude Code), `.cursor/mcp.json` (Cursor) | Claude: `mcpServers.coderifts` with `type: "http"` + `url`. Cursor: `url` only (no `type`). Never `httpUrl` (that is Gemini, extension-level — this command does not write it). |
+| **Agent rules** | generated host files via `agent-setup`'s writer | `AGENTS.md` is always written. Claude also gets `CLAUDE.md`; Cursor also gets `.cursor/rules/coderifts.mdc`. `--hosts=all` adds Copilot, LangGraph, and OpenAI instructions. (`coderifts init agent` is a different command — it writes a `.coderifts.yml` template.) |
+| **Host hook** | `.claude/settings.json`, `.cursor/hooks.json` | Reuses `hook install --claude` / `--cursor` (JSON-merge, backup, idempotent) |
+| **CI workflow** | `.github/workflows/coderifts.yml` | `coderifts/contract-gate@v0`. `require-verified-monitoring` is commented (honest default: **false**) |
+
+**Next manual step:** add `CODERIFTS_API_KEY` to the repository secrets. The
+CLI never writes a key anywhere.
+
+**This command does not:** enable branch protection (see
+[ENFORCEMENT.md](https://github.com/coderifts/contract-gate/blob/main/ENFORCEMENT.md)),
+make raw host tools inescapable, or enforce the workflow until
+`CodeRifts / contract-gate` is a required check in repo settings.
+
+The three separate commands this replaces for a fresh repo: `coderifts
+agent-setup`, `coderifts hook install --claude` (and `--cursor`), plus a
+hand-copied workflow file. Repo-level MCP config was generated by nothing
+before 4.6.0.
+
 ### `coderifts login`
 
 Save your CodeRifts API key for cloud features. Get a free key at [app.coderifts.com/api/signup](https://app.coderifts.com/api/signup).
@@ -101,13 +137,17 @@ Gate a **deploy** on the current `{ environment, artifact }` using a **signed**
 preflight receipt. Fail-closed by default (842 / hook pattern). The CD sibling of
 `publish-gate` / the merge-gate App check.
 
-The CLI **verifies locally** (Ed25519 against the key registry, kid status /
-retired-window, body-hash over the decision envelope, expiry with ID104
-leeway, intended-context bind of operation / environment / artifact /
-fingerprint). Crypto is the app kernel (`verifyReceipt` +
-`evaluateVerifyAuthorization`) — not reimplemented here. Then it binds the
-**computed** receipt view into the validated `deployGate` from
-`@coderifts/agent-guard`. No scoring is reimplemented.
+**Verification contract split (4.4.1 / guard 9.0.0):**
+
+- **Verdict:** `@coderifts/agent-guard` `deployGate` **TOKEN mode** is the
+  verifier of record. The CLI passes `{ token, decision_result, registry |
+  pinnedKeyPem }` — it never sends an unstamped `currently_authorized` view.
+- **WHY (842):** the CLI still runs the app kernel (`verifyReceipt` +
+  `evaluateVerifyAuthorization`) to corroborate `verify_status` and to print
+  envelope `blocking_reasons`. That is **not** a second verdict: one cause
+  (the gate `reason`), one WHY.
+
+Crypto is not reimplemented in the CLI. No scoring is reimplemented.
 
 **`--receipt` contract:** the file MUST carry the signed chain receipt token
 (`token` / `chain_receipt` / `receipt.token` / `decision_result.receipt.token`)

package/bin/coderifts.js

@@ -86,10 +86,49 @@ program
   });
 
 // ── init command ──
+// `init [template]` writes a policy YAML. `init --agents` is a different path:
+// it wires MCP config + agent rules + host hook + the CI gate workflow.
 program
   .command('init [template]')
-  .description('Generate a .coderifts.yml from a policy template (startup, growth, fintech, public-api, microservices)')
-  .action(async (template) => {
+  .description('Generate a .coderifts.yml from a policy template, or wire governed agent work (--agents)')
+  .option('--agents', 'Wire this repo for governed agent work (MCP config, rule files, host hook, CI gate)')
+  .option('--hosts <list>', 'With --agents: claude, cursor, or all (default: all)', 'all')
+  .option('--no-hook', 'With --agents: skip host hook install')
+  .option('--no-workflow', 'With --agents: skip writing .github/workflows/coderifts.yml')
+  .option('--dry-run', 'With --agents: print the plan without writing')
+  .option('--check', 'With --agents: report which of the four pieces are present (no write)')
+  .option('--out <dir>', 'With --agents: target directory (default: cwd)')
+  .action(async (template, options) => {
+    const {
+      runInitAgents,
+      agentFlagsRequireAgents,
+      AGENT_FLAGS_REQUIRE_AGENTS,
+      AGENTS_NO_TEMPLATE,
+    } = require('../src/commands/init-agents');
+    if (options.agents) {
+      if (template) {
+        console.error(AGENTS_NO_TEMPLATE);
+        process.exitCode = 1;
+        return;
+      }
+      const result = runInitAgents({
+        hosts: options.hosts,
+        hook: options.hook,
+        workflow: options.workflow,
+        dryRun: !!options.dryRun,
+        check: !!options.check,
+        out: options.out,
+      });
+      if (result && typeof result.exitCode === 'number') {
+        process.exitCode = result.exitCode;
+      }
+      return;
+    }
+    if (agentFlagsRequireAgents(options)) {
+      console.error(AGENT_FLAGS_REQUIRE_AGENTS);
+      process.exitCode = 1;
+      return;
+    }
     const { init } = require('../src/commands/init');
     await init(template);
   });
@@ -278,6 +317,27 @@ program
     }
   });
 
+// ── cursor-hook — Cursor preToolUse ({"permission":"deny"} on stdout; also exit 2) ──
+program
+  .command('cursor-hook')
+  .description(
+    'Cursor preToolUse gate: deny contract-touching Write|Delete on BLOCK/STOP/REQUEST_APPROVAL. '
+    + 'Emits {permission,user_message,agent_message} on stdout. Cursor does not enforce "ask" for '
+    + 'preToolUse, so REQUEST_APPROVAL maps to deny. Fail-closed on unparseable stdin.',
+  )
+  .addHelpText('after', () => {
+    const { USAGE } = require('../src/commands/cursor-hook');
+    return `\n${USAGE}\n`;
+  })
+  .action(async () => {
+    const { runCursorHook } = require('../src/commands/cursor-hook');
+    const fsMod = require('fs');
+    let stdin = '';
+    try { stdin = fsMod.readFileSync(0, 'utf8'); } catch { stdin = ''; }
+    const code = await runCursorHook({ argv: process.argv.slice(3), stdin });
+    if (typeof code === 'number') process.exitCode = code;
+  });
+
 // ── hook command group ──
 const hookCmd = program
   .command('hook')
@@ -287,12 +347,14 @@ hookCmd
   .command('install')
   .description('Install the CodeRifts pre-push hook in the current Git repo')
   .option('--claude', 'Write Claude Code PreToolUse hook to .claude/settings.json (JSON-merge)')
-  .option('--global', 'With --claude: write ~/.claude/settings.json instead of the project file')
-  .option('--log <path>', 'With --claude: embed CODERIFTS_GUARD_EVENT_LOG and CODERIFTS_SESSION_ID in the command')
+  .option('--cursor', 'Write Cursor preToolUse hook to .cursor/hooks.json (JSON-merge, failClosed:true)')
+  .option('--global', 'With --claude/--cursor: write the ~/ config instead of the project file')
+  .option('--log <path>', 'With --claude/--cursor: embed CODERIFTS_GUARD_EVENT_LOG and CODERIFTS_SESSION_ID in the command')
   .action((options) => {
     const { install } = require('../src/commands/hook');
     install({
       claude: !!options.claude,
+      cursor: !!options.cursor,
       global: !!options.global,
       log: options.log,
     });