@beryl-so/cli 0.5.0 → 0.7.1

package/README.md

@@ -26,6 +26,10 @@ npx @beryl-so/cli@latest …    # or run one-off without installing
 
 From the monorepo (development): `cd cli && npm install && npm run build && npm link`.
 
+`beryl init` and `beryl mcp` warn (stderr, best-effort) when the running CLI is behind the
+latest npm release — a stale MCP server silently exposes fewer tools. Set
+`BERYL_NO_UPDATE_CHECK=1` to opt out.
+
 ## Authenticate
 
 ```bash
@@ -73,6 +77,14 @@ Authenticate the CLI with your Beryl account
 | --- | --- | --- |
 | `beryl login` | Authenticate the CLI with your Beryl account | — |
 
+### signup
+
+Register a new Beryl account (emails a 6-digit verification code)
+
+| Command | Summary | MCP tool |
+| --- | --- | --- |
+| `beryl signup` | Register a new Beryl account (emails a 6-digit verification code) | `signup` |
+
 ### logout
 
 Remove the stored token from the CLI config
@@ -232,6 +244,7 @@ Inspect the agent's exploration runs — how it crawled a site and authored its
 | `beryl explorations list` | List the agent's exploration passes for a project | `explorations_list` |
 | `beryl explorations get <exploration-id>` | Show one exploration: authored tests, abandoned flows, coverage, frontier | `explorations_get` |
 | `beryl explorations steps <exploration-id>` | List every step the agent took in an exploration | `explorations_steps` |
+| `beryl explorations cancel <exploration-id>` | Cancel an in-flight exploration | `explorations_cancel` |
 | `beryl explorations watch <exploration-id>` | Stream an exploration live — watch the agent explore and author tests | `explorations_watch` |
 
 ### config
@@ -242,15 +255,31 @@ Manage the variables, secrets, and files the agent can use while exploring and r
 | --- | --- | --- |
 | `beryl config vars list` | List the project's config variables (visible to the agent during runs) | `config_vars_list` |
 | `beryl config vars set <key> <value>` | Create or update a config variable | `config_vars_set` |
+| `beryl config vars get <key>` | Show one config variable | `config_vars_get` |
 | `beryl config vars delete <key>` | Delete a config variable | `config_vars_delete` |
 | `beryl config secrets list` | List the project's secrets (values are never returned) | `config_secrets_list` |
 | `beryl config secrets set <key> <value>` | Create a secret (write-only; re-setting a key replaces it) | `config_secrets_set` |
+| `beryl config secrets get <key>` | Show one secret's metadata (the value is never returned) | `config_secrets_get` |
 | `beryl config secrets delete <key>` | Delete a secret | `config_secrets_delete` |
 | `beryl config files list` | List files uploaded for the agent to use (e.g. CSVs, upload fixtures) | `config_files_list` |
+| `beryl config files get <file>` | Show one uploaded file's metadata | `config_files_get` |
 | `beryl config files upload <file>` | Upload a file | `config_files_upload` |
 | `beryl config files download <file-id>` | Get a short-lived download URL for a file | `config_files_download` |
 | `beryl config files delete <file-id>` | Delete an uploaded file | `config_files_delete` |
 
+### slack
+
+Send run outcomes to a Slack channel via a per-project incoming webhook.
+
+`beryl slack` with no subcommand runs `slack show`.
+
+| Command | Summary | MCP tool |
+| --- | --- | --- |
+| `beryl slack show` | Show the project's Slack alert config (webhook is masked) | `slack_show` |
+| `beryl slack set-webhook <url>` | Set the Slack incoming-webhook URL run alerts post to | `slack_set_webhook` |
+| `beryl slack clear` | Remove the project's Slack webhook (stops all alerts) | `slack_clear` |
+| `beryl slack test` | Post a sample alert to the configured webhook | `slack_test` |
+
 ### credentials
 
 Manage saved logins Beryl reuses to test behind authentication, and attach them to projects.
@@ -273,10 +302,22 @@ Drive a browser session that captures a target-site login for Beryl to reuse.
 | Command | Summary | MCP tool |
 | --- | --- | --- |
 | `beryl auth-capture start` | Start a login-capture browser session for the project (non-interactive) | `auth_capture_start` |
+| `beryl auth-capture login <session-id>` | Log into the target site headlessly with credentials (no human at the browser) | `auth_capture_login` |
 | `beryl auth-capture capture <session-id>` | Capture the session after the user has logged in via the live-view URL | `auth_capture_capture` |
 | `beryl auth-capture refresh <session-id>` | Capture a refreshed session for a project whose login is expiring | `auth_capture_refresh` |
 | `beryl auth-capture release <session-id>` | Release a login-capture browser session without capturing | `auth_capture_release` |
 
+### inbox
+
+Disposable email inboxes for testing flows that send mail — signups, OTPs, receipts.
+
+| Command | Summary | MCP tool |
+| --- | --- | --- |
+| `beryl inbox create` | Mint a disposable email inbox that Beryl receives mail for | `inbox_create` |
+| `beryl inbox list` | List the workspace's inboxes, newest first | `inbox_list` |
+| `beryl inbox read <inbox-id>` | Read the latest email from an inbox (waits for one to arrive) | `inbox_read` |
+| `beryl inbox emails <inbox-id>` | List the emails an inbox has received | `inbox_emails` |
+
 ### account
 
 View and update your personal account profile.

package/dist/adapters/cli.js

@@ -1,10 +1,10 @@
-import fs from "node:fs";
 import { loadConfig } from "../config.js";
 import { createContext } from "../context.js";
 import { CliError, EXIT_OK, EXIT_USAGE, UsageError } from "../errors.js";
 import { ApiClient } from "../http.js";
 import { autoFormat, bold, cyan, dim } from "../output.js";
 import { commandGroups, findCommand, groupSummary } from "../registry/index.js";
+import { cliVersion } from "../version-check.js";
 import { mcpToolFor } from "./mcp.js";
 export const GLOBAL_FLAGS = [
     { name: "version", description: "Print the CLI version and exit", alias: "V" },
@@ -13,15 +13,6 @@ export const GLOBAL_FLAGS = [
     { name: "token", description: "Personal access token (overrides config/BERYL_API_KEY)", value: true },
     { name: "help", description: "Show help", alias: "h" },
 ];
-export function cliVersion() {
-    try {
-        const pkg = JSON.parse(fs.readFileSync(new URL("../../package.json", import.meta.url), "utf8"));
-        return pkg.version;
-    }
-    catch {
-        return "unknown";
-    }
-}
 export function parseArgv(spec, tokens) {
     const flags = {};
     const positional = [];
@@ -222,21 +213,47 @@ function groupHelp(group, specs) {
     lines.push("", `Run ${cyan(`beryl ${group} <subcommand> --help`)} for details.`);
     return lines.join("\n");
 }
+// Whether argv[i] is a recognised global flag, and how many tokens it spans (2 when it
+// takes its value from the next token, 1 for boolean/alias/inline `--name=value` forms).
+function globalFlagSpan(argv, i) {
+    const tok = argv[i];
+    const eq = tok.indexOf("=");
+    const name = (eq === -1 ? tok : tok.slice(0, eq)).replace(/^--?/, "");
+    const flag = GLOBAL_FLAGS.find((g) => (tok.startsWith("--") ? g.name === name : "alias" in g && g.alias === name));
+    if (!flag)
+        return 0;
+    return "value" in flag && flag.value && eq === -1 ? 2 : 1;
+}
 export async function runCli(argv) {
+    // Global flags are position-independent: collect command words by skipping over
+    // recognised global flags (leaving them in place for parseArgv), and stop only at a
+    // flag we don't recognise as global — that one belongs to the command.
     const words = [];
-    let rest = argv;
-    for (const tok of argv) {
-        if (tok.startsWith("-"))
+    const wordIndices = [];
+    for (let i = 0; i < argv.length; i++) {
+        const tok = argv[i];
+        if (tok === "--")
             break;
+        if (tok.startsWith("-") && tok !== "-") {
+            const span = globalFlagSpan(argv, i);
+            if (span === 0)
+                break;
+            i += span - 1;
+            continue;
+        }
         words.push(tok);
+        wordIndices.push(i);
     }
-    if (argv.includes("--version") || argv.includes("-V") || argv[0] === "version") {
+    let rest = argv;
+    let restWordIndices = wordIndices;
+    if (argv.includes("--version") || argv.includes("-V") || words[0] === "version") {
         process.stdout.write(cliVersion() + "\n");
         return EXIT_OK;
     }
     if (words[0] === "help") {
         words.shift();
         rest = [...words, "--help"];
+        restWordIndices = words.map((_, i) => i);
     }
     if (words.length === 0) {
         process.stdout.write(rootHelp() + "\n");
@@ -266,7 +283,10 @@ export async function runCli(argv) {
         return EXIT_USAGE;
     }
     const { spec } = found;
-    const tokens = rest.slice(found.consumed);
+    // Drop the consumed command words wherever they sit — global flags before or between
+    // them stay in place for parseArgv.
+    const consumedIndices = new Set(restWordIndices.slice(0, found.consumed));
+    const tokens = rest.filter((_, i) => !consumedIndices.has(i));
     let parsed;
     try {
         parsed = parseArgv(spec, tokens);

package/dist/adapters/mcp.js

@@ -4,7 +4,7 @@ import { CallToolRequestSchema, ListToolsRequestSchema, } from "@modelcontextpro
 import { createContext } from "../context.js";
 import { CliError } from "../errors.js";
 import { commands } from "../registry/index.js";
-import { cliVersion } from "./cli.js";
+import { cliVersion, warnIfStale } from "../version-check.js";
 export function toolName(spec) {
     return spec.name.replace(/ /g, "_").replace(/-/g, "_");
 }
@@ -72,6 +72,9 @@ function toInput(spec, params) {
     return { args, flags };
 }
 export async function serveMcp(baseCtx) {
+    // Fire-and-forget staleness warning: a stale MCP server silently exposes fewer
+    // tools, and stderr is the one channel a stdio MCP server can safely log to.
+    void warnIfStale(cliVersion(), (msg) => console.error(msg));
     const server = new Server({ name: "beryl", version: cliVersion() }, { capabilities: { tools: {} } });
     server.setRequestHandler(ListToolsRequestSchema, () => ({
         tools: mcpTools().map((spec) => ({

package/dist/beryl-test-skill.js

@@ -5,6 +5,20 @@
 // the CLI and another in the docs. Edit here; `init` writes it verbatim.
 export const BERYL_TEST_SKILL_FILENAME = "SKILL.md";
 export const BERYL_TEST_SKILL_DIR = "beryl-test";
+// The example plan the skill shows verbatim. Exported so the test suite lints it with
+// lintPlan — the skill's documented shape must pass `beryl tests lint` on the first try.
+export const BERYL_TEST_SKILL_EXAMPLE_PLAN = {
+    steps: [
+        { action: "goto", url: "https://app.example.com/pricing" },
+        {
+            action: "expect",
+            expect_kind: "have_text",
+            selector: "h1",
+            expect_text: "Pricing",
+        },
+        { action: "expect", expect_kind: "visible", selector: "text=Pro plan" },
+    ],
+};
 export const BERYL_TEST_SKILL = `---
 name: beryl-test
 description: Author durable, healable end-to-end tests for a web app with Beryl. Use when writing, running, or fixing a Beryl test locally with your own coding agent — drafting the plan over the Playwright MCP, writing the natural-language intent, and running it with \`beryl runs local\`.
@@ -33,6 +47,28 @@ assertion**, a strong **natural-language intent**, and the **local run-fix loop*
    - **at least one step is an \`expect\`** (a test that asserts nothing is not a test).
    Optional \`before\` / \`after\` arrays hold setup and teardown; \`after\` runs even when a
    main step fails, so a create/update/delete flow can clean up the record it made.
+
+   An \`expect\` step's shape is \`{action: "expect", expect_kind, selector, expect_text}\`:
+   - \`expect_kind\` (required) is one of \`visible\`, \`attached\`, \`hidden\`, \`checked\`,
+     \`enabled\`, \`disabled\`, \`have_text\`, \`have_value\`, \`have_url\`, \`have_title\`,
+     \`have_count\`, \`persisted\`, \`gone\`, \`count_delta\`.
+   - \`selector\` is required for every kind except the page-level \`have_url\` /
+     \`have_title\` (those assert on the page, not an element).
+   - \`expect_text\` carries the expected string for \`have_text\` / \`have_value\` (exact
+     match on the element) and \`have_url\` / \`have_title\` (substring match on the page).
+     There is no \`value\` field on an expect and no bare \`text\` / \`url\` kind.
+   - Two kinds take other fields instead: \`have_count\` needs \`expect_count\`, and
+     \`count_delta\` needs \`capture_ref\` + \`expect_delta\` (vs a baseline banked by an
+     earlier \`capture_count\` step).
+
+   A fully valid minimal plan ("the pricing page renders"):
+
+   \`\`\`json
+${JSON.stringify(BERYL_TEST_SKILL_EXAMPLE_PLAN, null, 2)
+    .split("\n")
+    .map((line) => `   ${line}`)
+    .join("\n")}
+   \`\`\`
 3. **Validate offline, then create:**
    \`\`\`
    beryl tests lint --file plan.json        # check the plan against the schema, no network
@@ -60,8 +96,15 @@ observable proof the flow worked. Get this right and everything else follows.
   "the URL is /pricing" passes even on a blank or broken page that never rendered.
   Reserve a URL-only assertion for when the URL *is* the outcome (a form that lands on
   \`/thank-you\`) and no distinctive destination content is available.
-- \`expect\` kinds: \`text\` (a string is visible), \`url\` (the URL contains a value), \`gone\`
-  (an element disappeared — e.g. a spinner, or the item you just deleted).
+- The usual outcome kinds: \`visible\` (the success element showed up), \`have_text\` (an
+  element's text matches), \`have_url\` (the URL contains a value), \`gone\` (an element
+  disappeared — e.g. a spinner, or the item you just deleted). §1 has the full
+  \`expect_kind\` list and the step shape.
+- **\`have_text\` is an EXACT full-text match on the selector's element** — asserting
+  \`have_text: "Documentation"\` on \`body\` fails, because \`body\`'s text includes all the
+  nav chrome. Target the specific element that carries the text (the \`h1\`, the toast),
+  or assert the page instead (\`have_title\`, which is a substring match). To check "this
+  string is visible somewhere", use \`expect_kind: "visible"\` with a \`text=…\` selector.
 - **If you can't name a success signal, the flow is not test-worthy.** Don't bank a test
   that verifies nothing. Explore a different flow instead.
 - **Don't work around a real app failure to make a test go green.** If the flow is

package/dist/commands/auth.js

@@ -54,7 +54,9 @@ export const authCommands = [
         summary: "Authenticate the CLI with your Beryl account",
         description: "Signs in with an emailed one-time code and mints a personal access token, " +
             "which is stored in the CLI config. Pass --token to use an existing token " +
-            "from Account → API tokens instead. In CI, prefer the BERYL_API_KEY environment variable.",
+            "from Account → API tokens instead. Pass --email plus --code (the 6 digits " +
+            "from the email, e.g. read from a `beryl inbox`) to complete the OTP flow " +
+            "without a prompt. In CI, prefer the BERYL_API_KEY environment variable.",
         interactive: true,
         flags: [
             { name: "token", type: "string", description: "Use an existing personal access token" },
@@ -64,26 +66,48 @@ export const authCommands = [
                 description: "Email for the one-time code sign-in (default: your last sign-in, or your git " +
                     "user.email — offered as the prompt default)",
             },
+            {
+                name: "code",
+                type: "string",
+                description: "The emailed 6-digit code — skips the prompt for non-interactive use " +
+                    "(requires --email; also skips sending a fresh code, so pair it with a " +
+                    "code already requested via `beryl login`, `beryl signup`, or the API)",
+            },
             {
                 name: "token-name",
                 type: "string",
                 description: "Name for the minted token (default: CLI on <hostname>)",
             },
         ],
-        examples: ["beryl login", "beryl login --token beryl_pat_…", "beryl login --email you@example.com"],
+        examples: [
+            "beryl login",
+            "beryl login --token beryl_pat_…",
+            "beryl login --email you@example.com",
+            "beryl login --email agent@example.com --code 123456 --json",
+        ],
         async run(ctx, input) {
             const apiUrl = ctx.client.baseUrl;
             let token = flagStr(input, "token");
             if (!token) {
+                const codeFlag = flagStr(input, "code")?.replace(/\s/g, "");
+                if (codeFlag !== undefined) {
+                    if (!/^\d{6}$/.test(codeFlag))
+                        throw new UsageError("--code must be the 6 digits from the email");
+                    if (!flagStr(input, "email"))
+                        throw new UsageError("--code requires --email");
+                }
                 const email = flagStr(input, "email") ?? (await promptForEmail(ctx));
                 if (!email.includes("@"))
                     throw new UsageError(`"${email}" is not an email address`);
                 const anon = new ApiClient(apiUrl);
-                await anon.post("/auth/request-login-otp", { email });
-                ctx.err(dim(`Sent a 6-digit code to ${email}`));
-                ctx.err(dim("Not arriving? Check spam; if you're new to Beryl, sign up at https://beryl.so first — " +
-                    "or use `beryl login --token` with a token from beryl.so → Account → API tokens."));
-                const code = await promptForOtpCode(ctx);
+                let code = codeFlag;
+                if (!code) {
+                    await anon.post("/auth/request-login-otp", { email });
+                    ctx.err(dim(`Sent a 6-digit code to ${email}`));
+                    ctx.err(dim("Not arriving? Check spam; if you're new to Beryl, sign up at https://beryl.so first — " +
+                        "or use `beryl login --token` with a token from beryl.so → Account → API tokens."));
+                    code = await promptForOtpCode(ctx);
+                }
                 const login = (await anon.post("/auth/verify-otp", { email, code }));
                 if (!login.access_token)
                     throw new CliError("Login did not return an access token");
@@ -125,6 +149,36 @@ export const authCommands = [
             };
         },
     },
+    {
+        name: "signup",
+        summary: "Register a new Beryl account (emails a 6-digit verification code)",
+        description: "Creates a passwordless account for the email and sends it a 6-digit code. " +
+            "Finish with `beryl login --email <addr> --code <the 6 digits>` — that verifies " +
+            "the account, creates its workspace, and signs the CLI in. With an inbox from " +
+            "`beryl inbox create` as the address, an agent can provision a fresh account " +
+            "end-to-end with no human at a prompt.",
+        flags: [
+            {
+                name: "email",
+                type: "string",
+                required: true,
+                description: "Email address for the new account",
+            },
+        ],
+        examples: ["beryl signup --email agent+ci@example.com --json"],
+        async run(ctx, input) {
+            const email = flagStr(input, "email") ?? "";
+            if (!email.includes("@"))
+                throw new UsageError(`"${email}" is not an email address`);
+            const anon = new ApiClient(ctx.client.baseUrl);
+            const created = (await anon.post("/auth/signup", { email }));
+            return {
+                data: created,
+                human: `${green("Signup started")} for ${email}\n` +
+                    dim(`Sent a 6-digit code to ${email} — finish with: beryl login --email ${email} --code <code>`),
+            };
+        },
+    },
     {
         name: "logout",
         summary: "Remove the stored token from the CLI config",

package/dist/commands/config-vars.js

@@ -3,13 +3,16 @@ import path from "node:path";
 import { UsageError } from "../errors.js";
 import { arg, flagBool, flagStr, projectPath } from "./util.js";
 const configPath = (ws, p) => `${projectPath(ws, p)}/config`;
-async function resolveByKey(rows, value, kind) {
+function findByKey(rows, value, kind) {
     const byId = rows.find((r) => r.id === value);
     if (byId)
-        return byId.id;
+        return byId;
     const byKey = rows.filter((r) => r.key === value);
     if (byKey.length === 1)
-        return byKey[0].id;
+        return byKey[0];
+    if (byKey.length > 1) {
+        throw new UsageError(`Multiple ${kind}s have key "${value}" (per-environment copies) — use the id instead`);
+    }
     throw new UsageError(`No ${kind} with key or id "${value}"`);
 }
 export const configCommands = [
@@ -51,17 +54,30 @@ export const configCommands = [
             };
         },
     },
+    {
+        name: "config vars get",
+        summary: "Show one config variable",
+        scope: "project",
+        args: [{ name: "key", description: "Variable key or id", required: true }],
+        async run(ctx, input) {
+            const { workspaceId, projectId } = await ctx.requireProject(input);
+            const rows = (await ctx.client.get(`${configPath(workspaceId, projectId)}/variables`));
+            return { data: findByKey(rows, arg(input, "key"), "variable") };
+        },
+    },
     {
         name: "config vars delete",
         summary: "Delete a config variable",
         scope: "project",
         args: [{ name: "key", description: "Variable key or id", required: true }],
+        flags: [{ name: "force", type: "boolean", description: "Skip the confirmation prompt" }],
         async run(ctx, input) {
             const { workspaceId, projectId } = await ctx.requireProject(input);
             const base = `${configPath(workspaceId, projectId)}/variables`;
             const rows = (await ctx.client.get(base));
-            const id = await resolveByKey(rows, arg(input, "key"), "variable");
-            await ctx.client.del(`${base}/${id}`);
+            const row = findByKey(rows, arg(input, "key"), "variable");
+            await ctx.confirm(`Delete config variable ${row.key}?`, flagBool(input, "force"));
+            await ctx.client.del(`${base}/${row.id}`);
             return { human: "Deleted." };
         },
     },
@@ -94,17 +110,30 @@ export const configCommands = [
             return { data: await ctx.client.post(base, { key, value, environment_id: envId }) };
         },
     },
+    {
+        name: "config secrets get",
+        summary: "Show one secret's metadata (the value is never returned)",
+        scope: "project",
+        args: [{ name: "key", description: "Secret key or id", required: true }],
+        async run(ctx, input) {
+            const { workspaceId, projectId } = await ctx.requireProject(input);
+            const rows = (await ctx.client.get(`${configPath(workspaceId, projectId)}/secrets`));
+            return { data: findByKey(rows, arg(input, "key"), "secret") };
+        },
+    },
     {
         name: "config secrets delete",
         summary: "Delete a secret",
         scope: "project",
         args: [{ name: "key", description: "Secret key or id", required: true }],
+        flags: [{ name: "force", type: "boolean", description: "Skip the confirmation prompt" }],
         async run(ctx, input) {
             const { workspaceId, projectId } = await ctx.requireProject(input);
             const base = `${configPath(workspaceId, projectId)}/secrets`;
             const rows = (await ctx.client.get(base));
-            const id = await resolveByKey(rows, arg(input, "key"), "secret");
-            await ctx.client.del(`${base}/${id}`);
+            const row = findByKey(rows, arg(input, "key"), "secret");
+            await ctx.confirm(`Delete secret ${row.key}?`, flagBool(input, "force"));
+            await ctx.client.del(`${base}/${row.id}`);
             return { human: "Deleted." };
         },
     },
@@ -117,6 +146,27 @@ export const configCommands = [
             return { data: await ctx.client.get(`${configPath(workspaceId, projectId)}/files`) };
         },
     },
+    {
+        name: "config files get",
+        summary: "Show one uploaded file's metadata",
+        scope: "project",
+        args: [{ name: "file", description: "File name or id", required: true }],
+        async run(ctx, input) {
+            const { workspaceId, projectId } = await ctx.requireProject(input);
+            const rows = (await ctx.client.get(`${configPath(workspaceId, projectId)}/files`));
+            const value = arg(input, "file");
+            const byId = rows.find((r) => r.id === value);
+            if (byId)
+                return { data: byId };
+            const byName = rows.filter((r) => r.name === value);
+            if (byName.length === 1)
+                return { data: byName[0] };
+            if (byName.length > 1) {
+                throw new UsageError(`Multiple files are named "${value}" (per-environment copies) — use the id instead`);
+            }
+            throw new UsageError(`No file with name or id "${value}"`);
+        },
+    },
     {
         name: "config files upload",
         summary: "Upload a file",
@@ -126,8 +176,18 @@ export const configCommands = [
         async run(ctx, input) {
             const { workspaceId, projectId } = await ctx.requireProject(input);
             const file = arg(input, "file");
+            let bytes;
+            try {
+                bytes = new Uint8Array(fs.readFileSync(file));
+            }
+            catch (err) {
+                const code = err.code;
+                if (code === "ENOENT")
+                    throw new UsageError(`file not found: ${file}`);
+                throw new UsageError(`cannot read ${file} (${code ?? err.message})`);
+            }
             const form = new FormData();
-            form.set("file", new Blob([fs.readFileSync(file)]), path.basename(file));
+            form.set("file", new Blob([bytes]), path.basename(file));
             const envId = flagStr(input, "env");
             if (envId)
                 form.set("environment_id", envId);

package/dist/commands/credentials.js

@@ -1,5 +1,6 @@
+import { UsageError } from "../errors.js";
 import { dim, green, yellow } from "../output.js";
-import { arg, flagBool } from "./util.js";
+import { arg, flagBool, flagStr } from "./util.js";
 const capturePath = (ws, p) => `/auth-capture/workspaces/${ws}/projects/${p}/sessions`;
 export const credentialCommands = [
     {
@@ -110,6 +111,37 @@ export const credentialCommands = [
             return { data: await ctx.client.post(capturePath(workspaceId, projectId)) };
         },
     },
+    {
+        name: "auth-capture login",
+        summary: "Log into the target site headlessly with credentials (no human at the browser)",
+        description: "Drives the login inside the capture session started by `auth-capture start`, so " +
+            "an agent can complete start → login → capture with zero human intervention. The " +
+            "credentials are sent to the server, typed into the target site over the wire, and " +
+            "never stored, logged, or returned — the captured session stays encrypted " +
+            "server-side. Follow with `auth-capture capture` to snapshot the authenticated session.",
+        scope: "project",
+        args: [
+            { name: "session-id", description: "Session id from auth-capture start", required: true },
+        ],
+        flags: [
+            { name: "username", type: "string", description: "Login username / email", required: true },
+            { name: "password", type: "string", description: "Login password", required: true },
+            {
+                name: "login-url",
+                type: "string",
+                description: "Explicit login page URL (defaults to the session's current page)",
+            },
+        ],
+        async run(ctx, input) {
+            const { workspaceId, projectId } = await ctx.requireProject(input);
+            const username = flagStr(input, "username");
+            const password = flagStr(input, "password");
+            if (!username || !password)
+                throw new UsageError("--username and --password are required");
+            await ctx.client.post(`${capturePath(workspaceId, projectId)}/${arg(input, "session-id")}/login`, { username, password, login_url: flagStr(input, "login-url") ?? null });
+            return { human: "Logged in." };
+        },
+    },
     {
         name: "auth-capture capture",
         summary: "Capture the session after the user has logged in via the live-view URL",

package/dist/commands/explorations.js

@@ -35,6 +35,20 @@ export const explorationCommands = [
             };
         },
     },
+    {
+        name: "explorations cancel",
+        summary: "Cancel an in-flight exploration",
+        description: "Stops a running exploration without touching the project or its tests — the " +
+            "lever for a runaway/non-convergent pass burning agent budget.",
+        scope: "project",
+        args: [{ name: "exploration-id", description: "Exploration id", required: true }],
+        async run(ctx, input) {
+            const { workspaceId, projectId } = await ctx.requireProject(input);
+            return {
+                data: await ctx.client.post(`${projectPath(workspaceId, projectId)}/explorations/${arg(input, "exploration-id")}/cancel`),
+            };
+        },
+    },
     {
         name: "explorations watch",
         summary: "Stream an exploration live — watch the agent explore and author tests",

package/dist/commands/inboxes.js

@@ -0,0 +1,129 @@
+import { dim, green } from "../output.js";
+import { arg, flagBool, flagNum, flagStr } from "./util.js";
+// Mirrors the API's own extractor (email_inbox/step_resolver.py): a labelled digit run
+// ("your code is 654321") beats the bare fenced pattern, because a real sign-in mail is
+// full of innocent 4-8 digit runs — "© 2026", a support number — and the bare fence
+// would happily return the first of them.
+const CODE_PATTERN = /(?<!\d)(\d{4,8})(?!\d)/;
+const LABELLED_CODE_PATTERN = /(?:verification|security|one[\s-]?time|login|sign[\s-]?in|access|confirmation)?\s*(?:code|otp|passcode|pin)\b[^0-9]{0,20}(?<!\d)(\d{4,8})(?!\d)/i;
+function visibleText(html) {
+    return html
+        .replace(/<(style|script|head)\b[\s\S]*?<\/\1>/gi, " ")
+        .replace(/<[^>]+>/g, " ");
+}
+function extractCode(email) {
+    const body = email.body_text || (email.body_html ? visibleText(email.body_html) : "");
+    for (const pattern of [LABELLED_CODE_PATTERN, CODE_PATTERN]) {
+        for (const text of [body, email.subject ?? ""]) {
+            const match = text.match(pattern);
+            if (match)
+                return match[1];
+        }
+    }
+    return null;
+}
+export const inboxCommands = [
+    {
+        name: "inbox create",
+        summary: "Mint a disposable email inbox that Beryl receives mail for",
+        groupSummary: "Disposable email inboxes for testing flows that send mail — signups, OTPs, receipts.",
+        description: "Creates a receiving address under Beryl's inbound email domain and returns it. " +
+            "Use it wherever a test needs a real, readable mailbox — e.g. as the --email for " +
+            "`beryl signup`, then read the code back with `beryl inbox read --extract-code`.",
+        scope: "workspace",
+        flags: [
+            {
+                name: "ttl-hours",
+                type: "number",
+                description: "Hours before the inbox expires and stops receiving (1-168, default 24)",
+            },
+            { name: "project", type: "string", description: "Attach the inbox to a project id" },
+        ],
+        examples: ["beryl inbox create --json", "beryl inbox create --ttl-hours 2"],
+        async run(ctx, input) {
+            const ws = await ctx.requireWorkspace(input);
+            const inbox = (await ctx.client.post(`/workspaces/${ws}/inboxes`, {
+                ttl_hours: flagNum(input, "ttl-hours") ?? 24,
+                project_id: flagStr(input, "project") ?? null,
+            }));
+            return {
+                data: inbox,
+                human: `${green("Created")} inbox ${inbox.id}\n\n  ${inbox.address}\n\n` +
+                    dim(`Read it with: beryl inbox read ${inbox.id}`),
+            };
+        },
+    },
+    {
+        name: "inbox list",
+        summary: "List the workspace's inboxes, newest first",
+        description: "Every inbox the workspace has minted with `beryl inbox create`. Expired inboxes " +
+            "stop receiving and are hard-deleted by a background sweep, so they drop off " +
+            "this list shortly after their TTL.",
+        scope: "workspace",
+        async run(ctx, input) {
+            const ws = await ctx.requireWorkspace(input);
+            return { data: await ctx.client.get(`/workspaces/${ws}/inboxes`) };
+        },
+    },
+    {
+        name: "inbox read",
+        summary: "Read the latest email from an inbox (waits for one to arrive)",
+        description: "Waits up to --timeout-s for a matching email and returns it (one blocking request; " +
+            "the server caps the wait at 50s — re-run to keep waiting). With --extract-code, " +
+            "also pulls the one-time code (4-8 digits) out of the body/subject — handy for " +
+            "completing `beryl login --email <addr> --code <code>` unattended. Exits non-zero " +
+            "if nothing arrives before the timeout.",
+        scope: "workspace",
+        args: [{ name: "inbox-id", description: "Inbox id from `beryl inbox create`", required: true }],
+        flags: [
+            {
+                name: "timeout-s",
+                type: "number",
+                description: "Seconds to wait for a matching email (0 = don't wait; max 50, default 30)",
+            },
+            { name: "since", type: "string", description: "Only emails received after this ISO timestamp" },
+            { name: "from-contains", type: "string", description: "Only emails whose sender contains this" },
+            {
+                name: "subject-contains",
+                type: "string",
+                description: "Only emails whose subject contains this",
+            },
+            {
+                name: "extract-code",
+                type: "boolean",
+                description: "Also return the one-time code found in the email as `code`",
+            },
+        ],
+        examples: [
+            "beryl inbox read ibx_123 --timeout-s 45 --json",
+            "beryl inbox read ibx_123 --subject-contains code --extract-code --json",
+        ],
+        async run(ctx, input) {
+            const ws = await ctx.requireWorkspace(input);
+            const email = (await ctx.client.get(`/workspaces/${ws}/inboxes/${arg(input, "inbox-id")}/emails/latest`, {
+                timeout_s: flagNum(input, "timeout-s"),
+                since: flagStr(input, "since"),
+                from_contains: flagStr(input, "from-contains"),
+                subject_contains: flagStr(input, "subject-contains"),
+            }));
+            if (!flagBool(input, "extract-code"))
+                return { data: email };
+            return { data: { ...email, code: extractCode(email) } };
+        },
+    },
+    {
+        name: "inbox emails",
+        summary: "List the emails an inbox has received",
+        scope: "workspace",
+        args: [{ name: "inbox-id", description: "Inbox id from `beryl inbox create`", required: true }],
+        flags: [
+            { name: "since", type: "string", description: "Only emails received after this ISO timestamp" },
+        ],
+        async run(ctx, input) {
+            const ws = await ctx.requireWorkspace(input);
+            return {
+                data: await ctx.client.get(`/workspaces/${ws}/inboxes/${arg(input, "inbox-id")}/emails`, { since: flagStr(input, "since") }),
+            };
+        },
+    },
+];

package/dist/commands/init.js

@@ -4,12 +4,17 @@ import { BERYL_TEST_SKILL, BERYL_TEST_SKILL_DIR, BERYL_TEST_SKILL_FILENAME, } fr
 import { LOCAL_CONFIG_FILENAME, loadConfig } from "../config.js";
 import { CliError, UsageError } from "../errors.js";
 import { ApiClient } from "../http.js";
-import { bold, cyan, dim, green } from "../output.js";
+import { bold, cyan, dim, green, yellow } from "../output.js";
+import { cliVersion, warnIfStale } from "../version-check.js";
 import { authCommands } from "./auth.js";
 import { flagBool, flagStr } from "./util.js";
+// Pinned to @latest (like the Playwright entry below) — an unpinned spec lets npx
+// serve whatever stale global/cache install already resolves, so the agent's MCP
+// server would silently miss newer tools. mergeMcpConfig compares entries verbatim,
+// so re-running init also upgrades an old unpinned entry in place.
 const MCP_SERVER_ENTRY = {
     command: "npx",
-    args: ["-y", "@beryl-so/cli", "mcp"],
+    args: ["-y", "@beryl-so/cli@latest", "mcp"],
 };
 const PLAYWRIGHT_SERVER_ENTRY = {
     command: "npx",
@@ -263,6 +268,9 @@ export const initCommands = [
                         ? `\n\nPlaywright MCP is wired — your coding agent can explore the site locally ` +
                             `and push tests with ${cyan("beryl tests create")}.`
                         : "");
+            // The .mcp.json entry is pinned to @latest, but a global install / old npx cache
+            // still wins resolution — so tell the user when the CLI they just ran is stale.
+            await warnIfStale(cliVersion(), (msg) => ctx.err(yellow(msg)));
             return {
                 data: { workspace: workspaceId ?? null, project: projectId ?? null, editors },
                 human: nextSteps,

package/dist/commands/slack.js

@@ -0,0 +1,82 @@
+import { arg, flagBool, projectPath } from "./util.js";
+const slackPath = (ws, p) => `${projectPath(ws, p)}/slack`;
+export const slackCommands = [
+    {
+        name: "slack show",
+        summary: "Show the project's Slack alert config (webhook is masked)",
+        groupSummary: "Send run outcomes to a Slack channel via a per-project incoming webhook.",
+        scope: "project",
+        groupDefault: true,
+        async run(ctx, input) {
+            const { workspaceId, projectId } = await ctx.requireProject(input);
+            return { data: await ctx.client.get(slackPath(workspaceId, projectId)) };
+        },
+    },
+    {
+        name: "slack set-webhook",
+        summary: "Set the Slack incoming-webhook URL run alerts post to",
+        scope: "project",
+        args: [
+            {
+                name: "url",
+                description: "Slack incoming-webhook URL (https://hooks.slack.com/…)",
+                required: true,
+            },
+        ],
+        flags: [
+            {
+                name: "all-runs",
+                type: "boolean",
+                description: "Post on every run, not only on a new regression",
+            },
+            {
+                name: "failures-only",
+                type: "boolean",
+                description: "Post only on a new regression (the default)",
+            },
+        ],
+        async run(ctx, input) {
+            const { workspaceId, projectId } = await ctx.requireProject(input);
+            const path = slackPath(workspaceId, projectId);
+            // Preserve the stored notify preference unless the caller states one, so
+            // rotating the webhook doesn't silently reset an existing all-runs opt-in.
+            let notifyAllRuns;
+            if (flagBool(input, "all-runs")) {
+                notifyAllRuns = true;
+            }
+            else if (flagBool(input, "failures-only")) {
+                notifyAllRuns = false;
+            }
+            else {
+                const current = (await ctx.client.get(path));
+                notifyAllRuns = current?.notify_all_runs ?? false;
+            }
+            return {
+                data: await ctx.client.put(path, {
+                    webhook_url: arg(input, "url"),
+                    notify_all_runs: notifyAllRuns,
+                }),
+            };
+        },
+    },
+    {
+        name: "slack clear",
+        summary: "Remove the project's Slack webhook (stops all alerts)",
+        scope: "project",
+        async run(ctx, input) {
+            const { workspaceId, projectId } = await ctx.requireProject(input);
+            await ctx.client.del(slackPath(workspaceId, projectId));
+            return { human: "Cleared." };
+        },
+    },
+    {
+        name: "slack test",
+        summary: "Post a sample alert to the configured webhook",
+        scope: "project",
+        async run(ctx, input) {
+            const { workspaceId, projectId } = await ctx.requireProject(input);
+            await ctx.client.post(`${slackPath(workspaceId, projectId)}/test`);
+            return { human: "Sent a test message to Slack." };
+        },
+    },
+];

package/dist/commands/tests.js

@@ -1,9 +1,23 @@
 import fs from "node:fs";
 import { UsageError } from "../errors.js";
 import { lintPlan } from "../lint.js";
+import { table } from "../output.js";
 import { ACTION_PLAN_SCHEMA } from "../schema.generated.js";
 import { arg, argList, flagBool, flagNum, flagStr, projectPath, readJsonFlag } from "./util.js";
 const testPath = (ws, p, id) => `${projectPath(ws, p)}/tests/${id}`;
+// The concise human table for `tests list` — the full TestResponse is a 24-column
+// firehose of internal ids that wraps unreadably in a normal terminal. `--wide`
+// (and `--json`) still expose every field.
+const LIST_COLUMNS = ["title", "status", "id", "last_result_status", "last_run_at"];
+function conciseTestRow(row) {
+    return {
+        title: row.nl_title,
+        status: row.quarantined ? "quarantined" : row.is_active ? "active" : "inactive",
+        id: row.id,
+        last_result_status: row.last_result_status,
+        last_run_at: row.last_run_at,
+    };
+}
 export const testCommands = [
     {
         name: "tests lint",
@@ -33,17 +47,24 @@ export const testCommands = [
     {
         name: "tests list",
         summary: "List the project's tests with their latest result",
+        description: "Prints a concise table by default (title / status / id / last result / last run). " +
+            "Pass --wide for every field, or --json for the raw records.",
         scope: "project",
         groupDefault: true,
         groupSummary: "Author, inspect, version, and heal a project's tests — the checks Beryl runs on each run.",
-        flags: [{ name: "env", type: "string", description: "Filter by environment id" }],
+        flags: [
+            { name: "env", type: "string", description: "Filter by environment id" },
+            { name: "wide", type: "boolean", description: "Show all columns, not the concise default" },
+        ],
         async run(ctx, input) {
             const { workspaceId, projectId } = await ctx.requireProject(input);
-            return {
-                data: await ctx.client.get(`${projectPath(workspaceId, projectId)}/tests`, {
-                    environment_id: flagStr(input, "env"),
-                }),
-            };
+            const data = (await ctx.client.get(`${projectPath(workspaceId, projectId)}/tests`, {
+                environment_id: flagStr(input, "env"),
+            }));
+            if (flagBool(input, "wide"))
+                return { data };
+            // `data` stays the full records so --json is unchanged; only the human table is trimmed.
+            return { data, human: table(data.map(conciseTestRow), LIST_COLUMNS) };
         },
     },
     {

package/dist/commands/util.js

@@ -29,11 +29,24 @@ export function flagNum(input, name) {
         throw new UsageError(`--${name} must be a number`);
     return n;
 }
+/** Read a local file for a flag/arg, turning fs errors (ENOENT, EACCES, EISDIR…) into a
+ *  clean usage message instead of a raw Node stack. */
+export function readFileArg(path, label) {
+    try {
+        return fs.readFileSync(path === "-" ? 0 : path, "utf8");
+    }
+    catch (err) {
+        const code = err.code;
+        if (code === "ENOENT")
+            throw new UsageError(`${label}: file not found: ${path}`);
+        throw new UsageError(`${label}: cannot read ${path} (${code ?? err.message})`);
+    }
+}
 export function readJsonFlag(input, name) {
     const file = flagStr(input, name);
     if (!file)
         throw new UsageError(`--${name} <file> is required`);
-    const text = file === "-" ? fs.readFileSync(0, "utf8") : fs.readFileSync(file, "utf8");
+    const text = readFileArg(file, `--${name}`);
     try {
         return JSON.parse(text);
     }

package/dist/registry/index.js

@@ -4,10 +4,12 @@ import { configCommands } from "../commands/config-vars.js";
 import { credentialCommands } from "../commands/credentials.js";
 import { environmentCommands } from "../commands/environments.js";
 import { explorationCommands } from "../commands/explorations.js";
+import { inboxCommands } from "../commands/inboxes.js";
 import { initCommands } from "../commands/init.js";
 import { mcpCommands } from "../commands/mcp.js";
 import { projectCommands } from "../commands/projects.js";
 import { runCommands } from "../commands/runs.js";
+import { slackCommands } from "../commands/slack.js";
 import { testCommands } from "../commands/tests.js";
 import { workspaceCommands } from "../commands/workspaces.js";
 export const WORKSPACE_FLAG = {
@@ -43,7 +45,9 @@ export const commands = [
     ...runCommands,
     ...explorationCommands,
     ...configCommands,
+    ...slackCommands,
     ...credentialCommands,
+    ...inboxCommands,
     ...accountCommands,
     ...mcpCommands,
 ].map(withScopeFlags);

package/dist/version-check.js

@@ -0,0 +1,65 @@
+import fs from "node:fs";
+// Best-effort "are we stale?" check against the npm registry. The `.mcp.json` entry
+// `init` writes pins `@beryl-so/cli@latest`, but a user with a global install (or an
+// old npx cache) can still be running a stale CLI — and an MCP server that's behind
+// silently exposes fewer tools. So `init` and `beryl mcp` warn when the running
+// version is behind the registry's `latest`. Never fatal: offline, a slow registry,
+// or an unparseable response just skips the warning.
+const REGISTRY_LATEST_URL = "https://registry.npmjs.org/@beryl-so/cli/latest";
+const TIMEOUT_MS = 1500;
+export function cliVersion() {
+    try {
+        const pkg = JSON.parse(fs.readFileSync(new URL("../package.json", import.meta.url), "utf8"));
+        return pkg.version;
+    }
+    catch {
+        return "unknown";
+    }
+}
+// Numeric-triple compare; returns true only when `current` is strictly behind
+// `latest`. Anything non-numeric (dev builds, "unknown", dist-tags) compares as
+// not-behind so we never nag on unparseable versions.
+export function isBehind(current, latest) {
+    const parse = (v) => {
+        const m = /^(\d+)\.(\d+)\.(\d+)/.exec(v.trim());
+        return m ? [Number(m[1]), Number(m[2]), Number(m[3])] : null;
+    };
+    const cur = parse(current);
+    const lat = parse(latest);
+    if (!cur || !lat)
+        return false;
+    for (let i = 0; i < 3; i++) {
+        if (cur[i] !== lat[i])
+            return cur[i] < lat[i];
+    }
+    return false;
+}
+/** The registry's `latest` version, or null when it can't be fetched quickly. */
+export async function fetchLatestVersion() {
+    try {
+        const res = await fetch(REGISTRY_LATEST_URL, {
+            signal: AbortSignal.timeout(TIMEOUT_MS),
+        });
+        if (!res.ok)
+            return null;
+        const body = (await res.json());
+        return typeof body.version === "string" ? body.version : null;
+    }
+    catch {
+        return null;
+    }
+}
+/**
+ * Warn (via `warn`) when `current` is behind the registry's `latest`.
+ * Best-effort and quiet on any failure — safe to await on every startup.
+ * BERYL_NO_UPDATE_CHECK=1 opts out entirely (no network call).
+ */
+export async function warnIfStale(current, warn) {
+    if (process.env.BERYL_NO_UPDATE_CHECK)
+        return;
+    const latest = await fetchLatestVersion();
+    if (latest && isBehind(current, latest)) {
+        warn(`@beryl-so/cli ${current} is behind the latest release ${latest} — ` +
+            `some commands/MCP tools may be missing. Upgrade with: npm i -g @beryl-so/cli@latest`);
+    }
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@beryl-so/cli",
-  "version": "0.5.0",
+  "version": "0.7.1",
   "description": "Beryl on the command line — projects, runs, the exploring agent, and an MCP server over the same commands.",
   "license": "MIT",
   "type": "module",