firegoat 0.5.0 → 0.7.0

package/dist/cli.js

@@ -6947,7 +6947,7 @@ var require_public_api = __commonJS(function(exports) {
 // package.json
 var package_default = {
   name: "firegoat",
-  version: "0.5.0",
+  version: "0.7.0",
   repository: {
     type: "git",
     url: "git+https://github.com/team2027/firegoat-cli.git"
@@ -6987,9 +6987,411 @@ var package_default = {
   packageManager: "bun@1.4.3"
 };
 
+// src/claude-login.ts
+import { spawnSync as spawnSync2 } from "node:child_process";
+import { mkdirSync as mkdirSync3 } from "node:fs";
+
+// src/auth.ts
+import {
+  chmodSync,
+  existsSync as existsSync2,
+  mkdirSync as mkdirSync2,
+  readFileSync,
+  writeFileSync
+} from "node:fs";
+import { join as join2 } from "node:path";
+
+// src/sandbox.ts
+import { spawnSync } from "node:child_process";
+import { existsSync, mkdirSync, realpathSync } from "node:fs";
+import { homedir } from "node:os";
+import { basename, dirname, join } from "node:path";
+var SANDBOX_EXEC = "/usr/bin/sandbox-exec";
+function firegoatHome() {
+  return process.env.FIREGOAT_HOME || join(homedir(), ".firegoat");
+}
+function claudeConfigDir() {
+  return join(firegoatHome(), "claude");
+}
+var TOOLCHAINS = [
+  ".nvm",
+  ".fnm",
+  ".volta",
+  ".bun",
+  ".deno",
+  ".cargo",
+  ".rustup",
+  ".pyenv",
+  ".rbenv",
+  ".asdf",
+  ".sdkman",
+  ".local/share/mise",
+  ".local/share/fnm",
+  ".local/share/claude",
+  ".local/share/uv",
+  "go",
+  "Library/pnpm"
+];
+var TOOLCHAIN_SECRETS = [".cargo/credentials", ".cargo/credentials.toml"];
+function real(p) {
+  return realpathSync(p);
+}
+function under(p, dir) {
+  return p === dir || p.startsWith(`${dir}/`);
+}
+function readableToolDirs(home, claudeBin) {
+  const dirs = new Set;
+  const add = (p) => {
+    if (existsSync(p))
+      dirs.add(real(p));
+  };
+  for (const t of TOOLCHAINS)
+    add(join(home, t));
+  for (const p of (process.env.PATH ?? "").split(":")) {
+    if (!p.startsWith("/") || !existsSync(p))
+      continue;
+    const r = real(p);
+    if (!under(r, home) || r === home)
+      continue;
+    const prefix = dirname(r);
+    dirs.add(basename(r) === "bin" && prefix !== home && prefix !== join(home, ".local") ? prefix : r);
+  }
+  add(dirname(dirname(process.execPath)));
+  add(dirname(claudeBin));
+  add(dirname(real(claudeBin)));
+  return [...dirs];
+}
+var q = (s) => JSON.stringify(s);
+function sandboxProfile(opts) {
+  const home = real(homedir());
+  const reads = readableToolDirs(home, opts.claudeBin).map((d) => `(subpath ${q(d)})`);
+  return `(version 1)
+(allow default)
+
+(deny file-read* (subpath ${q(home)}))
+(allow file-read-metadata)
+(allow file-read*
+  ${reads.join(`
+  `)}
+  (subpath ${q(join(home, "Library/Keychains"))})
+  (regex ${q(`^${home}/Library/Preferences/com\\.apple\\.security`)})
+  (regex ${q(`^${home}/Library/Preferences/\\.GlobalPreferences`)}))
+(deny file-read* file-write*
+  ${TOOLCHAIN_SECRETS.map((s) => `(literal ${q(join(home, s))})`).join(`
+  `)})
+
+(deny file-write* (subpath "/"))
+(allow file-write*
+  (subpath "/dev")
+  (subpath "/private/tmp")
+  (subpath "/private/var/folders"))
+(allow file-read* file-write*
+  (subpath ${q(real(opts.configDir))})
+  (subpath ${q(real(opts.runRoot))}))
+
+(deny file-read* file-write* (subpath ${q(real(opts.repo))}))
+`;
+}
+function agentEnv(runRoot, apiKey) {
+  const env = { ...process.env };
+  delete env.ANTHROPIC_API_KEY;
+  if (apiKey)
+    env.ANTHROPIC_API_KEY = apiKey;
+  const cache = join(runRoot, "cache");
+  const tmp = join(runRoot, "tmp");
+  mkdirSync(cache, { recursive: true });
+  mkdirSync(tmp, { recursive: true });
+  return {
+    ...env,
+    CLAUDE_CONFIG_DIR: claudeConfigDir(),
+    TMPDIR: tmp,
+    XDG_CACHE_HOME: cache,
+    npm_config_cache: join(cache, "npm"),
+    npm_config_store_dir: join(cache, "pnpm-store"),
+    YARN_CACHE_FOLDER: join(cache, "yarn"),
+    BUN_INSTALL_CACHE_DIR: join(cache, "bun"),
+    PIP_CACHE_DIR: join(cache, "pip"),
+    UV_CACHE_DIR: join(cache, "uv"),
+    GOMODCACHE: join(cache, "gomod"),
+    GOCACHE: join(cache, "gobuild"),
+    CARGO_HOME: join(cache, "cargo")
+  };
+}
+function wrap(profile, cmd, args) {
+  if (profile === null)
+    return { cmd, args };
+  return { cmd: SANDBOX_EXEC, args: ["-p", profile, cmd, ...args] };
+}
+function probeSandbox(opts) {
+  const outside = join(homedir(), `.firegoat-escape-probe-${process.pid}-${Date.now()}`);
+  const script = 'touch "$1" 2>/dev/null && echo home-write; cat "$2" >/dev/null 2>&1 && echo repo-read; touch "$3" 2>/dev/null && echo work-write; rm -f "$1" "$3"';
+  const { cmd, args } = wrap(opts.profile, "/bin/sh", [
+    "-c",
+    script,
+    "probe",
+    outside,
+    opts.repoFile,
+    join(opts.workdir, ".firegoat-probe")
+  ]);
+  const r = spawnSync(cmd, args, { encoding: "utf8", cwd: opts.workdir });
+  if (r.error)
+    return `sandbox probe could not start: ${r.error.message}`;
+  const got = r.stdout.trim().split(`
+`).filter(Boolean);
+  if (got.includes("home-write") || got.includes("repo-read"))
+    return `sandbox does not hold (escaped: ${got.filter((g) => g !== "work-write").join(", ")}); refusing to run the agent`;
+  if (!got.includes("work-write"))
+    return `sandbox blocks writes in the workdir (stderr: ${r.stderr.trim() || "none"})`;
+  return;
+}
+
+// src/auth.ts
+var authFile = () => join2(firegoatHome(), "auth.json");
+var keyFile = () => join2(firegoatHome(), "anthropic-api-key");
+
+class AuthError extends Error {
+}
+function readMethod() {
+  const file = authFile();
+  if (!existsSync2(file))
+    return;
+  let doc;
+  try {
+    doc = JSON.parse(readFileSync(file, "utf8"));
+  } catch (e) {
+    throw new AuthError(`${file} is not valid JSON: ${e.message}`);
+  }
+  const method = doc?.method;
+  if (method !== "claude" && method !== "api-key")
+    throw new AuthError(`${file}: method must be "claude" or "api-key", got ${JSON.stringify(method)}. Run: npx firegoat init`);
+  return method;
+}
+function saveMethod(method) {
+  mkdirSync2(firegoatHome(), { recursive: true });
+  writeFileSync(authFile(), `${JSON.stringify({ method }, null, 2)}
+`);
+}
+function saveApiKey(apiKey) {
+  mkdirSync2(firegoatHome(), { recursive: true });
+  const file = keyFile();
+  writeFileSync(file, `${apiKey}
+`, { mode: 384 });
+  chmodSync(file, 384);
+  saveMethod("api-key");
+  return file;
+}
+function looksLikeAnthropicKey(s) {
+  return /^sk-ant-[A-Za-z0-9_-]{20,}$/.test(s);
+}
+function resolveAuth(env) {
+  if (env.CI && env.ANTHROPIC_API_KEY)
+    return {
+      method: "api-key",
+      label: "Anthropic API key from ANTHROPIC_API_KEY (CI)",
+      apiKey: env.ANTHROPIC_API_KEY
+    };
+  const method = readMethod() ?? "claude";
+  if (method === "claude")
+    return {
+      method,
+      label: `Claude login in ${claudeConfigDir()}`
+    };
+  const file = keyFile();
+  if (!existsSync2(file))
+    throw new AuthError(`auth is set to api-key but ${file} is missing. Run: npx firegoat init`);
+  const apiKey = readFileSync(file, "utf8").trim();
+  if (!apiKey)
+    throw new AuthError(`${file} is empty. Run: npx firegoat init`);
+  return { method, label: `Anthropic API key in ${file}`, apiKey };
+}
+
+// src/exit.ts
+var EXIT = { pass: 0, fail: 1, notGraded: 2, usage: 3 };
+
+// src/help.ts
+var EXIT_CODES = `Exit codes:
+  0  every run's script checks passed
+  1  a script check failed
+  2  at least one run was not graded (claude missing or logged out, crash, timeout,
+     setup failed); wins over 1
+  3  usage or config error (bad flag, unknown test, invalid or missing firegoat.yaml)`;
+function topHelp(version) {
+  return `firegoat ${version}: local AX evals. A coding agent gets one prompt in a fresh,
+sandboxed folder; firegoat records the transcript and runs your script checks.
+
+Typical flow (for an agent driving firegoat on behalf of a human):
+  1. npx firegoat init --yes          write firegoat.yaml (read: npx firegoat init --help)
+  2. edit firegoat.yaml               real user problems as prompts, a goal, checks
+  3. ask the human to run once:       npx firegoat claude login   (opens a browser)
+  4. npx firegoat eval --json         run the tests; JSON summary on stdout
+  5. judge each run's goal + rubric yourself, from its run.json, transcript and
+     workdir: firegoat only runs the script checks
+
+Commands:
+  init           write firegoat.yaml; choose how the agent signs in
+  claude login   sign claude in to firegoat's own config dir (interactive, once)
+  eval           run the tests
+  open           open (or --print) a run's folder
+
+Run firegoat <command> --help for details and examples.
+
+${EXIT_CODES}`;
+}
+var INIT_HELP = `Usage: firegoat init [--yes] [--force] [--skills] [--auth claude|api-key]
+
+Writes firegoat.yaml in the current folder with starter tests from what the repo
+ships (package.json, pyproject.toml, go.mod, Cargo.toml, README), and adds
+.firegoat/ to .gitignore. Edit the tests afterwards; they are templates.
+
+  --yes, -y       don't ask questions (default when stdin is not a terminal)
+  --force, -f     overwrite an existing firegoat.yaml
+  --skills        also run: npx skills add team2027/skills --skill firegoat-onboarding
+                  --skill suggesting-eval-prompts (into .claude/skills/)
+  --auth claude   the agent signs in with a Claude login: runs \`firegoat claude login\`
+                  (interactive: a human must complete it in the browser)
+  --auth api-key  the agent signs in with an Anthropic API key, read from stdin when
+                  piped (else asked for without echo) and stored in
+                  ~/.firegoat/anthropic-api-key (mode 0600), never in the repo
+                  Without --auth, init asks (unless --yes). In CI, ANTHROPIC_API_KEY in
+                  the env wins.
+
+firegoat.yaml reference:
+  agents: [claude-code]       only claude-code for now
+  runs: 1                     runs per test
+  setup: <sh> | [<sh>, ...]   once per eval, in the repo, on the host; whatever it
+                              writes to $FIREGOAT_OUT is copied into each run and put
+                              first on the agent's PATH
+  workdir_setup: <sh> | [...] in each fresh workdir before the agent starts
+                              (env: $FIREGOAT_OUT, $FIREGOAT_WORKDIR, vars)
+  vars: {KEY: value}          defaults for {{vars.KEY}}; --var KEY=value overrides
+  tests:
+    - name: quickstart        letters, digits, '.', '_', '-'; unique
+      prompt: ...             what a user would ask the agent; one concrete outcome
+      goal: ...               one sentence that is true when the problem is solved
+      rubric: [...]           optional criteria for whoever grades the run (you)
+      checks:                 optional shell commands, run in the workdir after the
+        - run: ...            agent; exit 0 = pass; they set eval's exit code
+      timeout: 600            optional, seconds
+
+  The agent's folder has no repo files. Checks get $FIREGOAT_WORKDIR,
+  $FIREGOAT_SESSION (stream-json transcript), $FIREGOAT_OUT and the vars.
+  goal and rubric are not run by firegoat: they are handed to the caller to judge.
+
+Example:
+  setup: npm pack --pack-destination "$FIREGOAT_OUT"
+  workdir_setup: npm init -y >/dev/null && npm install "$FIREGOAT_OUT"/*.tgz
+  tests:
+    - name: first-call
+      prompt: >-
+        Using the acme SDK (already installed from the build under test), send a test
+        event and print the event id. Credentials are in ACME_API_KEY. Keep it minimal.
+      goal: A script sends one event and prints its id.
+      rubric:
+        - uses the SDK's client, not raw HTTP
+        - does not install acme from the registry
+      checks:
+        - run: grep -qE 'evt_[A-Za-z0-9]+' "$FIREGOAT_SESSION"
+
+Next: npx firegoat claude login (a human, once), then npx firegoat eval.
+
+Exit codes: 0 done, 2 the skills install or claude login failed, 3 usage error.`;
+var EVAL_HELP = `Usage: firegoat eval [test..] [--runs N] [--var KEY=value].. [--json] [--host]
+                     [--claude-bin PATH]
+
+Runs each test headless: claude gets the prompt in a fresh folder outside the repo,
+sandboxed on macOS (it can't read your home dir or the repo), with setup:'s outputs
+on PATH. Then the test's script checks run in that folder.
+
+  test..             only these tests (by name)
+  --runs N           runs per test (overrides runs: in firegoat.yaml)
+  --var KEY=value    fills {{vars.KEY}}; env for setup, workdir_setup and checks
+  --json             the summary as JSON on stdout; progress goes to stderr
+  --host             no sandbox (required off macOS; use a throwaway container)
+  --claude-bin PATH  claude binary (env FIREGOAT_CLAUDE_BIN; default: claude on PATH)
+
+Sign-in: the one chosen in firegoat init (Claude login or stored API key); in CI,
+ANTHROPIC_API_KEY from the env. eval prints which. Not logged in = exit 2 with the
+command to run.
+
+Output: .firegoat/runs/<runId>/
+  summary.json            every run, plus totals and a grading hint
+  <test>/<n>/run.json     one run: test, n, status and reason (from the checks),
+                          prompt, goal, rubric, checks [{run, code}], and paths:
+                          dir, workdir, session, stderr, checksLog
+  <test>/<n>/session.jsonl  the agent's transcript, stream-json (one event per line)
+  setup.txt, <test>/<n>/workdir_setup.txt, checks.txt   logs
+
+Grading: firegoat runs only the script checks; they set the exit code. The goal and
+rubric are yours to judge, with your own process. session.jsonl is stream-json: the
+last line has type "result" with the agent's final message; tool calls are
+"tool_use" items in "assistant" messages; tool output is in "user" messages.
+
+Examples:
+  npx firegoat eval --json                  all tests, JSON summary on stdout
+  npx firegoat eval quickstart --runs 3     one test, three runs
+  npx firegoat eval --var SHA=$(git rev-parse HEAD)
+
+firegoat.yaml is looked up in the current folder, then parents up to the git root.
+A run can take minutes (default timeout 600s per test).
+
+${EXIT_CODES}`;
+var OPEN_HELP = `Usage: firegoat open [runId] [--print]
+
+Opens a run's folder (.firegoat/runs/<runId>/) with the system opener; the latest
+run by default. --print only prints the path (use this from an agent).
+
+Exit codes: 0 done, 2 the opener failed, 3 no such run.`;
+var CLAUDE_LOGIN_HELP = `Usage: firegoat claude login [--claude-bin PATH]
+
+Signs claude in to firegoat's own config dir (~/.firegoat/claude), once per machine.
+Evals then run on that login, never your ~/.claude. Interactive: it opens a browser,
+so an agent should ask the human to run it. \`firegoat login\` is a deprecated alias.
+
+Exit codes: 0 logged in, 2 claude missing or the login failed, 3 usage error.`;
+
+// src/claude-login.ts
+function claudeLoginCommand(args) {
+  let claudeBin = process.env.FIREGOAT_CLAUDE_BIN || "claude";
+  for (let i = 0;i < args.length; i++) {
+    const a = args[i];
+    if (a === "--help" || a === "-h") {
+      console.log(CLAUDE_LOGIN_HELP);
+      return EXIT.pass;
+    }
+    if (a === "--claude-bin" && args[i + 1] !== undefined)
+      claudeBin = args[++i];
+    else if (a.startsWith("--claude-bin="))
+      claudeBin = a.slice(13);
+    else {
+      console.error(`firegoat claude login: unknown argument: ${a}`);
+      return EXIT.usage;
+    }
+  }
+  const configDir = claudeConfigDir();
+  mkdirSync3(configDir, { recursive: true });
+  const env = {
+    ...process.env,
+    CLAUDE_CONFIG_DIR: configDir
+  };
+  delete env.ANTHROPIC_API_KEY;
+  console.error(`Signing claude in to ${configDir}. Evals run with this login, never your own ~/.claude.`);
+  const r = spawnSync2(claudeBin, ["auth", "login"], { env, stdio: "inherit" });
+  if (r.error) {
+    console.error(`firegoat claude login: could not run ${claudeBin}: ${r.error.message}`);
+    return EXIT.notGraded;
+  }
+  if (r.status !== 0) {
+    console.error(`firegoat claude login: claude auth login exited ${r.status}`);
+    return EXIT.notGraded;
+  }
+  saveMethod("claude");
+  console.log("logged in; evals now sign in with this Claude login. next: npx firegoat eval");
+  return EXIT.pass;
+}
+
 // src/config.ts
-import { existsSync, readFileSync } from "node:fs";
-import { dirname, join } from "node:path";
+import { existsSync as existsSync3, readFileSync as readFileSync2 } from "node:fs";
+import { dirname as dirname2, join as join3 } from "node:path";
 
 // node_modules/yaml/dist/index.js
 var composer = require_composer();
@@ -7062,7 +7464,7 @@ function posInt(v, path, fallback) {
     fail(path, "must be a positive integer");
   return v;
 }
-function parseConfig(text, warn = console.error) {
+function parseConfig(text) {
   let raw;
   try {
     raw = $parse(text);
@@ -7102,20 +7504,21 @@ function parseConfig(text, warn = console.error) {
     if (!Array.isArray(checksRaw))
       fail(`${p}.checks`, "must be a list");
     const checks = [];
+    const rubric = rubricOf(t.rubric, `${p}.rubric`);
     checksRaw.forEach((c, j) => {
       const cp = `${p}.checks[${j}]`;
       if (!isRecord(c))
         fail(cp, "must be a mapping");
-      if ("judge" in c) {
-        warn(`${CONFIG_FILE}: ${cp}: judge checks need upload, which v1 does not do; skipped`);
-        return;
-      }
-      checks.push({ run: str(c.run, `${cp}.run`) });
+      if ("judge" in c)
+        rubric.push(str(c.judge, `${cp}.judge`));
+      else
+        checks.push({ run: str(c.run, `${cp}.run`) });
     });
     return {
       name,
       prompt: str(t.prompt, `${p}.prompt`),
       goal: str(t.goal, `${p}.goal`),
+      rubric,
       checks,
       timeout: posInt(t.timeout, `${p}.timeout`, DEFAULT_TIMEOUT)
     };
@@ -7130,6 +7533,15 @@ function parseConfig(text, warn = console.error) {
   };
 }
 var NAME = /^[A-Za-z0-9._-]+$/;
+function rubricOf(raw, path) {
+  if (raw === undefined)
+    return [];
+  if (typeof raw === "string")
+    return [str(raw, path)];
+  if (!Array.isArray(raw))
+    fail(path, "must be a string or a list of strings");
+  return raw.map((r, i) => str(r, `${path}[${i}]`));
+}
 function commands(raw, path) {
   if (raw === undefined)
     return [];
@@ -7167,19 +7579,19 @@ function render(text, vars) {
     return v;
   });
 }
-function loadConfig(dir, warn) {
+function loadConfig(dir) {
   const file = `${dir}/${CONFIG_FILE}`;
-  if (!existsSync(file))
+  if (!existsSync3(file))
     throw new ConfigError(`${CONFIG_FILE} not found in ${dir}. Run: npx firegoat init`);
-  return parseConfig(readFileSync(file, "utf8"), warn);
+  return parseConfig(readFileSync2(file, "utf8"));
 }
 function findConfigDir(cwd) {
   let dir = cwd;
   for (;; ) {
-    if (existsSync(join(dir, CONFIG_FILE)))
+    if (existsSync3(join3(dir, CONFIG_FILE)))
       return dir;
-    const parent = dirname(dir);
-    if (existsSync(join(dir, ".git")) || parent === dir)
+    const parent = dirname2(dir);
+    if (existsSync3(join3(dir, ".git")) || parent === dir)
       return cwd;
     dir = parent;
   }
@@ -7189,164 +7601,20 @@ function findConfigDir(cwd) {
 import { randomBytes } from "node:crypto";
 import {
   createWriteStream as createWriteStream3,
-  existsSync as existsSync3,
-  mkdirSync as mkdirSync3,
+  existsSync as existsSync4,
+  mkdirSync as mkdirSync5,
   mkdtempSync,
-  readFileSync as readFileSync2,
+  readFileSync as readFileSync3,
   realpathSync as realpathSync2,
-  writeFileSync
+  writeFileSync as writeFileSync2
 } from "node:fs";
 import { tmpdir } from "node:os";
-import { join as join5, resolve as resolve2 } from "node:path";
+import { join as join6, resolve as resolve2 } from "node:path";
 
 // src/claude.ts
-import { spawn, spawnSync as spawnSync2 } from "node:child_process";
+import { spawn, spawnSync as spawnSync3 } from "node:child_process";
 import { createWriteStream } from "node:fs";
 import { createInterface } from "node:readline";
-
-// src/sandbox.ts
-import { spawnSync } from "node:child_process";
-import { existsSync as existsSync2, mkdirSync, realpathSync } from "node:fs";
-import { homedir } from "node:os";
-import { basename, dirname as dirname2, join as join2 } from "node:path";
-var SANDBOX_EXEC = "/usr/bin/sandbox-exec";
-function firegoatHome() {
-  return process.env.FIREGOAT_HOME || join2(homedir(), ".firegoat");
-}
-function claudeConfigDir() {
-  return join2(firegoatHome(), "claude");
-}
-var TOOLCHAINS = [
-  ".nvm",
-  ".fnm",
-  ".volta",
-  ".bun",
-  ".deno",
-  ".cargo",
-  ".rustup",
-  ".pyenv",
-  ".rbenv",
-  ".asdf",
-  ".sdkman",
-  ".local/share/mise",
-  ".local/share/fnm",
-  ".local/share/claude",
-  ".local/share/uv",
-  "go",
-  "Library/pnpm"
-];
-var TOOLCHAIN_SECRETS = [".cargo/credentials", ".cargo/credentials.toml"];
-function real(p) {
-  return realpathSync(p);
-}
-function under(p, dir) {
-  return p === dir || p.startsWith(`${dir}/`);
-}
-function readableToolDirs(home, claudeBin) {
-  const dirs = new Set;
-  const add = (p) => {
-    if (existsSync2(p))
-      dirs.add(real(p));
-  };
-  for (const t of TOOLCHAINS)
-    add(join2(home, t));
-  for (const p of (process.env.PATH ?? "").split(":")) {
-    if (!p.startsWith("/") || !existsSync2(p))
-      continue;
-    const r = real(p);
-    if (!under(r, home) || r === home)
-      continue;
-    const prefix = dirname2(r);
-    dirs.add(basename(r) === "bin" && prefix !== home && prefix !== join2(home, ".local") ? prefix : r);
-  }
-  add(dirname2(dirname2(process.execPath)));
-  add(dirname2(claudeBin));
-  add(dirname2(real(claudeBin)));
-  return [...dirs];
-}
-var q = (s) => JSON.stringify(s);
-function sandboxProfile(opts) {
-  const home = real(homedir());
-  const reads = readableToolDirs(home, opts.claudeBin).map((d) => `(subpath ${q(d)})`);
-  return `(version 1)
-(allow default)
-
-(deny file-read* (subpath ${q(home)}))
-(allow file-read-metadata)
-(allow file-read*
-  ${reads.join(`
-  `)}
-  (subpath ${q(join2(home, "Library/Keychains"))})
-  (regex ${q(`^${home}/Library/Preferences/com\\.apple\\.security`)})
-  (regex ${q(`^${home}/Library/Preferences/\\.GlobalPreferences`)}))
-(deny file-read* file-write*
-  ${TOOLCHAIN_SECRETS.map((s) => `(literal ${q(join2(home, s))})`).join(`
-  `)})
-
-(deny file-write* (subpath "/"))
-(allow file-write*
-  (subpath "/dev")
-  (subpath "/private/tmp")
-  (subpath "/private/var/folders"))
-(allow file-read* file-write*
-  (subpath ${q(real(opts.configDir))})
-  (subpath ${q(real(opts.runRoot))}))
-
-(deny file-read* file-write* (subpath ${q(real(opts.repo))}))
-`;
-}
-function agentEnv(runRoot) {
-  const env = { ...process.env };
-  delete env.ANTHROPIC_API_KEY;
-  const cache = join2(runRoot, "cache");
-  const tmp = join2(runRoot, "tmp");
-  mkdirSync(cache, { recursive: true });
-  mkdirSync(tmp, { recursive: true });
-  return {
-    ...env,
-    CLAUDE_CONFIG_DIR: claudeConfigDir(),
-    TMPDIR: tmp,
-    XDG_CACHE_HOME: cache,
-    npm_config_cache: join2(cache, "npm"),
-    npm_config_store_dir: join2(cache, "pnpm-store"),
-    YARN_CACHE_FOLDER: join2(cache, "yarn"),
-    BUN_INSTALL_CACHE_DIR: join2(cache, "bun"),
-    PIP_CACHE_DIR: join2(cache, "pip"),
-    UV_CACHE_DIR: join2(cache, "uv"),
-    GOMODCACHE: join2(cache, "gomod"),
-    GOCACHE: join2(cache, "gobuild"),
-    CARGO_HOME: join2(cache, "cargo")
-  };
-}
-function wrap(profile, cmd, args) {
-  if (profile === null)
-    return { cmd, args };
-  return { cmd: SANDBOX_EXEC, args: ["-p", profile, cmd, ...args] };
-}
-function probeSandbox(opts) {
-  const outside = join2(homedir(), `.firegoat-escape-probe-${process.pid}-${Date.now()}`);
-  const script = 'touch "$1" 2>/dev/null && echo home-write; cat "$2" >/dev/null 2>&1 && echo repo-read; touch "$3" 2>/dev/null && echo work-write; rm -f "$1" "$3"';
-  const { cmd, args } = wrap(opts.profile, "/bin/sh", [
-    "-c",
-    script,
-    "probe",
-    outside,
-    opts.repoFile,
-    join2(opts.workdir, ".firegoat-probe")
-  ]);
-  const r = spawnSync(cmd, args, { encoding: "utf8", cwd: opts.workdir });
-  if (r.error)
-    return `sandbox probe could not start: ${r.error.message}`;
-  const got = r.stdout.trim().split(`
-`).filter(Boolean);
-  if (got.includes("home-write") || got.includes("repo-read"))
-    return `sandbox does not hold (escaped: ${got.filter((g) => g !== "work-write").join(", ")}); refusing to run the agent`;
-  if (!got.includes("work-write"))
-    return `sandbox blocks writes in the workdir (stderr: ${r.stderr.trim() || "none"})`;
-  return;
-}
-
-// src/claude.ts
 var CLAUDE_FLAGS = [
   "--output-format",
   "stream-json",
@@ -7511,8 +7779,8 @@ async function runClaude(opts) {
           badStart = `claude output did not start with a system/init stream-json line (got: ${line.slice(0, 120)})`;
           return "kill";
         }
-        if (msg.apiKeySource !== "none") {
-          badStart = `claude would use API key billing (apiKeySource=${JSON.stringify(msg.apiKeySource)}) instead of your Claude subscription. firegoat removes ANTHROPIC_API_KEY from the agent's env; check apiKeyHelper or other API key settings, then run \`claude /login\``;
+        if (msg.apiKeySource !== opts.keySource) {
+          badStart = opts.keySource === "none" ? `claude would use API key billing (apiKeySource=${JSON.stringify(msg.apiKeySource)}) instead of your Claude login. firegoat removes ANTHROPIC_API_KEY from the agent's env; check apiKeyHelper or other API key settings, then run \`npx firegoat claude login\`` : `claude did not use the configured API key (apiKeySource=${JSON.stringify(msg.apiKeySource)}, expected "ANTHROPIC_API_KEY")`;
           return "kill";
         }
         return;
@@ -7592,7 +7860,7 @@ async function runShell(opts) {
 }
 function checkLogin(opts) {
   const { cmd, args } = wrap(opts.profile, opts.bin, ["auth", "status"]);
-  const r = spawnSync2(cmd, args, {
+  const r = spawnSync3(cmd, args, {
     cwd: opts.cwd,
     env: opts.env,
     encoding: "utf8",
@@ -7607,16 +7875,13 @@ function checkLogin(opts) {
   if (status?.loggedIn === true)
     return;
   if (status?.loggedIn === false)
-    return `claude is not logged in to firegoat's config dir (${opts.env.CLAUDE_CONFIG_DIR}). Run: npx firegoat login`;
+    return `claude is not logged in to firegoat's config dir (${opts.env.CLAUDE_CONFIG_DIR}). Run: npx firegoat claude login`;
   return `claude auth status gave no login state (exit ${r.status}): ${(r.stdout + r.stderr).trim().slice(0, 300)}`;
 }
 
-// src/exit.ts
-var EXIT = { pass: 0, fail: 1, notGraded: 2, usage: 3 };
-
 // src/setup.ts
-import { cpSync, createWriteStream as createWriteStream2, mkdirSync as mkdirSync2 } from "node:fs";
-import { delimiter, join as join3 } from "node:path";
+import { cpSync, createWriteStream as createWriteStream2, mkdirSync as mkdirSync4 } from "node:fs";
+import { delimiter, join as join4 } from "node:path";
 var SETUP_TIMEOUT_MS = 600000;
 async function runAll(commands, cwd, env, logPath, what) {
   const log = createWriteStream2(logPath);
@@ -7638,13 +7903,13 @@ async function runAll(commands, cwd, env, logPath, what) {
   }
 }
 async function setupOnce(opts) {
-  mkdirSync2(opts.out, { recursive: true });
+  mkdirSync4(opts.out, { recursive: true });
   return runAll(opts.commands, opts.root, { ...opts.env, FIREGOAT_OUT: opts.out }, opts.logPath, "setup");
 }
 async function prepareRun(opts) {
-  const out = join3(opts.runRoot, "out");
+  const out = join4(opts.runRoot, "out");
   if (opts.out === null)
-    mkdirSync2(out);
+    mkdirSync4(out);
   else
     cpSync(opts.out, out, { recursive: true });
   const path = [out, opts.env.PATH ?? ""].join(delimiter);
@@ -7659,7 +7924,7 @@ async function prepareRun(opts) {
 
 // src/which.ts
 import { accessSync, constants, statSync } from "node:fs";
-import { delimiter as delimiter2, join as join4, resolve } from "node:path";
+import { delimiter as delimiter2, join as join5, resolve } from "node:path";
 function resolveBin(bin) {
   const executable = (p) => {
     try {
@@ -7672,14 +7937,39 @@ function resolveBin(bin) {
   if (bin.includes("/"))
     return executable(bin) ? resolve(bin) : undefined;
   for (const dir of (process.env.PATH ?? "").split(delimiter2)) {
-    if (dir && executable(join4(dir, bin)))
-      return join4(dir, bin);
+    if (dir && executable(join5(dir, bin)))
+      return join5(dir, bin);
   }
   return;
 }
 
 // src/eval.ts
 var CHECK_TIMEOUT_MS = 120000;
+var GRADING_HINT = "firegoat ran only the script checks. The goal and rubric are for the caller to judge, from each run's session (stream-json transcript) and workdir.";
+function blankRun(test, n, dir, prompt) {
+  return {
+    test: test.name,
+    n,
+    status: "error",
+    reason: "",
+    prompt,
+    goal: test.goal,
+    rubric: test.rubric,
+    checks: [],
+    dir,
+    workdir: null,
+    session: join6(dir, "session.jsonl"),
+    stderr: join6(dir, "stderr.txt"),
+    checksLog: join6(dir, "checks.txt"),
+    num_turns: null,
+    cost: null,
+    duration_ms: 0
+  };
+}
+function writeRunJson(r) {
+  writeFileSync2(join6(r.dir, "run.json"), `${JSON.stringify(r, null, 2)}
+`);
+}
 
 class UsageError extends Error {
 }
@@ -7735,16 +8025,16 @@ function newRunId() {
   return `${ts}-${randomBytes(2).toString("hex")}`;
 }
 function gitignoreHasFiregoat(cwd) {
-  const file = join5(cwd, ".gitignore");
-  if (!existsSync3(file))
+  const file = join6(cwd, ".gitignore");
+  if (!existsSync4(file))
     return false;
-  return readFileSync2(file, "utf8").split(`
+  return readFileSync3(file, "utf8").split(`
 `).map((l) => l.trim()).some((l) => /^\/?\.firegoat\/?(\*\*?)?$/.test(l));
 }
 function newRunRoot(label) {
-  const runRoot = realpathSync2(mkdtempSync(join5(tmpdir(), `firegoat-${label}-`)));
-  const workdir = join5(runRoot, "work");
-  mkdirSync3(workdir);
+  const runRoot = realpathSync2(mkdtempSync(join6(tmpdir(), `firegoat-${label}-`)));
+  const workdir = join6(runRoot, "work");
+  mkdirSync5(workdir);
   return { runRoot, workdir };
 }
 function isolate(iso, claudeBin, runRoot, workdir) {
@@ -7762,25 +8052,15 @@ function isolate(iso, claudeBin, runRoot, workdir) {
 async function runOne(opts) {
   const { test, n } = opts;
   const started = Date.now();
-  const dir = join5(opts.runDir, test.name, String(n));
-  mkdirSync3(dir, { recursive: true });
-  const session = join5(dir, "session.jsonl");
-  const summary = {
-    test: test.name,
-    n,
-    status: "error",
-    reason: "",
-    workdir: null,
-    session,
-    checks: [],
-    num_turns: null,
-    cost: null,
-    duration_ms: 0
-  };
+  const dir = join6(opts.runDir, test.name, String(n));
+  mkdirSync5(dir, { recursive: true });
+  const summary = blankRun(test, n, dir, render(test.prompt, opts.vars));
+  const session = summary.session;
   const done = (status, reason) => {
     summary.status = status;
     summary.reason = reason;
     summary.duration_ms = Date.now() - started;
+    writeRunJson(summary);
     return summary;
   };
   const { runRoot, workdir } = newRunRoot(`${test.name}-${n}`);
@@ -7797,7 +8077,7 @@ async function runOne(opts) {
       workdir,
       commands: opts.workdirSetup,
       env: hostEnv,
-      logPath: join5(dir, "workdir_setup.txt")
+      logPath: join6(dir, "workdir_setup.txt")
     });
     if ("error" in prepared)
       return done("error", `not graded: ${prepared.error}`);
@@ -7806,13 +8086,17 @@ async function runOne(opts) {
   }
   const agentPath = extra.PATH ? { PATH: extra.PATH } : {};
   const outcome = await runClaude({
-    env: { ...agentEnv(runRoot), ...agentPath },
+    env: {
+      ...agentEnv(runRoot, opts.auth.method === "api-key" ? opts.auth.apiKey : undefined),
+      ...agentPath
+    },
+    keySource: opts.auth.method === "api-key" ? "ANTHROPIC_API_KEY" : "none",
     profile: sandbox.profile,
     bin: opts.claudeBin,
-    prompt: render(test.prompt, opts.vars),
+    prompt: summary.prompt,
     cwd: workdir,
     sessionPath: session,
-    stderrPath: join5(dir, "stderr.txt"),
+    stderrPath: summary.stderr,
     timeoutMs: test.timeout * 1000
   });
   if (outcome.result) {
@@ -7821,11 +8105,9 @@ async function runOne(opts) {
   }
   if (!outcome.ok)
     return done("error", `not graded: ${outcome.reason}`);
-  if (test.checks.length === 0) {
-    opts.warn(`warning: ${test.name} has no checks, so its goal is unverified; counted as pass`);
-    return done("pass", "no checks (goal unverified)");
-  }
-  const log = createWriteStream3(join5(dir, "checks.txt"));
+  if (test.checks.length === 0)
+    return done("pass", "no script checks; the goal is for the grader");
+  const log = createWriteStream3(summary.checksLog);
   const env = {
     ...process.env,
     ...hostEnv,
@@ -7847,7 +8129,7 @@ async function runOne(opts) {
   } finally {
     await new Promise((r) => log.end(r));
   }
-  summary.checks = results.map((r) => ({ run: r.run, code: r.code }));
+  summary.checks = results.map((r) => r.error === undefined ? { run: r.run, code: r.code } : { run: r.run, code: r.code, error: r.error });
   const broken = results.find((r) => r.notStarted);
   if (broken)
     return done("error", `not graded: check ${broken.run}: ${broken.error}`);
@@ -7857,6 +8139,10 @@ async function runOne(opts) {
   return done("pass", `${results.length}/${results.length} checks passed`);
 }
 async function evalCommand(args, cwd) {
+  if (args.includes("--help") || args.includes("-h")) {
+    console.log(EVAL_HELP);
+    return EXIT.pass;
+  }
   let opts;
   try {
     opts = parseArgs(args);
@@ -7870,7 +8156,7 @@ async function evalCommand(args, cwd) {
   const say = opts.json ? console.error : console.log;
   const warn = console.error;
   const root = findConfigDir(cwd);
-  const config = loadConfig(root, warn);
+  const config = loadConfig(root);
   let tests = config.tests;
   if (opts.names.length > 0) {
     const known = new Set(tests.map((t) => t.name));
@@ -7897,45 +8183,58 @@ async function evalCommand(args, cwd) {
   }
   if (opts.host)
     warn("warning: --host: the agent runs without isolation and can read and write your files");
-  else if (process.platform !== "darwin" || !existsSync3(SANDBOX_EXEC)) {
+  else if (process.platform !== "darwin" || !existsSync4(SANDBOX_EXEC)) {
     console.error(`firegoat eval: isolation needs macOS (${SANDBOX_EXEC}); this is ${process.platform}. In CI or Linux, run inside a throwaway container and pass --host.`);
     return EXIT.usage;
   }
   const iso = {
     host: opts.host,
     repo: root,
-    repoFile: join5(root, "firegoat.yaml")
+    repoFile: join6(root, "firegoat.yaml")
   };
-  mkdirSync3(claudeConfigDir(), { recursive: true });
+  let auth;
+  try {
+    auth = resolveAuth(process.env);
+  } catch (e) {
+    if (e instanceof AuthError) {
+      console.error(`firegoat eval: ${e.message}`);
+      return EXIT.usage;
+    }
+    throw e;
+  }
+  say(`auth: ${auth.label}`);
+  if (auth.method === "claude" && process.env.ANTHROPIC_API_KEY)
+    warn("note: ANTHROPIC_API_KEY in your shell is not passed to the agent (it is used only in CI, or after `firegoat init --auth api-key`)");
+  mkdirSync5(claudeConfigDir(), { recursive: true });
   const pre = newRunRoot("preflight");
   const preSandbox = isolate(iso, claudePath, pre.runRoot, pre.workdir);
-  const loginProblem = "error" in preSandbox ? preSandbox.error : checkLogin({
+  const loginProblem = "error" in preSandbox ? preSandbox.error : auth.method === "claude" ? checkLogin({
     bin: claudePath,
     env: agentEnv(pre.runRoot),
     profile: preSandbox.profile,
     cwd: pre.workdir
-  });
+  }) : undefined;
   if (loginProblem) {
     console.error(`firegoat eval: not graded: ${loginProblem}`);
     return EXIT.notGraded;
   }
-  mkdirSync3(join5(root, ".firegoat"), { recursive: true });
+  mkdirSync5(join6(root, ".firegoat"), { recursive: true });
   if (!gitignoreHasFiregoat(root))
     warn("hint: add .firegoat/ to .gitignore (it holds run traces)");
   const runId = newRunId();
-  const runDir = join5(root, ".firegoat", "runs", runId);
-  mkdirSync3(runDir, { recursive: true });
-  const summaryPath = join5(runDir, "summary.json");
+  const runDir = join6(root, ".firegoat", "runs", runId);
+  mkdirSync5(runDir, { recursive: true });
+  const summaryPath = join6(runDir, "summary.json");
   let out = null;
   if (config.setup.length > 0) {
     say("setup...");
-    out = join5(runDir, "out");
+    out = join6(runDir, "out");
     const error = await setupOnce({
       commands: config.setup,
       root,
       out,
       env: { ...process.env, ...vars },
-      logPath: join5(runDir, "setup.txt")
+      logPath: join6(runDir, "setup.txt")
     });
     if (error) {
       console.error(`firegoat eval: not graded: ${error}`);
@@ -7951,10 +8250,12 @@ async function evalCommand(args, cwd) {
       runId,
       runsDir: runDir,
       isolation: opts.host ? "host" : "seatbelt",
+      auth: auth.method,
+      grading: GRADING_HINT,
       totals,
       runs: results
     };
-    writeFileSync(summaryPath, `${JSON.stringify(doc, null, 2)}
+    writeFileSync2(summaryPath, `${JSON.stringify(doc, null, 2)}
 `);
     return doc;
   };
@@ -7968,24 +8269,18 @@ async function evalCommand(args, cwd) {
           runDir,
           claudeBin: claudePath,
           iso,
+          auth,
           out,
           workdirSetup: config.workdirSetup,
           vars,
           warn
         });
       } catch (e) {
-        r = {
-          test: test.name,
-          n,
-          status: "error",
-          reason: `not graded: firegoat crashed: ${e.message}`,
-          workdir: null,
-          session: join5(runDir, test.name, String(n), "session.jsonl"),
-          checks: [],
-          num_turns: null,
-          cost: null,
-          duration_ms: 0
-        };
+        const dir = join6(runDir, test.name, String(n));
+        mkdirSync5(dir, { recursive: true });
+        r = blankRun(test, n, dir, test.prompt);
+        r.reason = `not graded: firegoat crashed: ${e.message}`;
+        writeRunJson(r);
       }
       results.push(r);
       writeSummary();
@@ -7997,23 +8292,18 @@ async function evalCommand(args, cwd) {
   say(`
 ${pass} passed, ${fail} failed, ${error} not graded (${results.length} runs)`);
   say(`runs: ${runDir}`);
+  say(`goal and rubric are not graded by firegoat; per run: ${join6(runDir, "<test>", "<n>", "run.json")}`);
   if (opts.json)
     console.log(JSON.stringify(doc, null, 2));
   return error > 0 ? EXIT.notGraded : fail > 0 ? EXIT.fail : EXIT.pass;
 }
 
 // src/init.ts
-import { spawnSync as spawnSync3 } from "node:child_process";
-import { existsSync as existsSync4, readFileSync as readFileSync3, writeFileSync as writeFileSync2 } from "node:fs";
-import { basename as basename2, join as join6 } from "node:path";
+import { spawnSync as spawnSync4 } from "node:child_process";
+import { existsSync as existsSync5, readFileSync as readFileSync4, writeFileSync as writeFileSync3 } from "node:fs";
+import { basename as basename2, join as join7 } from "node:path";
 import { createInterface as createInterface2 } from "node:readline/promises";
-var INIT_USAGE = `Usage: firegoat init [--yes] [--force] [--skills]
-
-  --yes, -y     don't ask questions (default when stdin is not a terminal)
-  --force, -f   overwrite an existing ${CONFIG_FILE}
-  --skills      install the firegoat-onboarding and suggesting-eval-prompts skills
-                into .claude/skills/ (npx skills add team2027/skills)
-`;
+import { Writable } from "node:stream";
 var SKILLS = ["firegoat-onboarding", "suggesting-eval-prompts"];
 var SKILLS_ARGV = [
   "skills",
@@ -8027,7 +8317,7 @@ var SKILLS_ARGV = [
 var GITIGNORE_LINE = ".firegoat/";
 function readText(file) {
   try {
-    return readFileSync3(file, "utf8");
+    return readFileSync4(file, "utf8");
   } catch {
     return;
   }
@@ -8053,9 +8343,9 @@ function tomlSections(text) {
   return sections;
 }
 function detect(cwd) {
-  const readme = ["README.md", "README", "readme.md", "Readme.md"].find((f) => existsSync4(join6(cwd, f)));
+  const readme = ["README.md", "README", "readme.md", "Readme.md"].find((f) => existsSync5(join7(cwd, f)));
   const fallback = { name: basename2(cwd), library: true, readme };
-  const pkgText = readText(join6(cwd, "package.json"));
+  const pkgText = readText(join7(cwd, "package.json"));
   if (pkgText !== undefined) {
     try {
       const pkg = JSON.parse(pkgText);
@@ -8070,7 +8360,7 @@ function detect(cwd) {
       return { ecosystem: "npm", name, bin, library, readme, docsUrl };
     } catch {}
   }
-  const pyText = readText(join6(cwd, "pyproject.toml"));
+  const pyText = readText(join7(cwd, "pyproject.toml"));
   if (pyText !== undefined) {
     const s = tomlSections(pyText);
     const name = s.get("project")?.get("name") ?? s.get("tool.poetry")?.get("name") ?? fallback.name;
@@ -8080,12 +8370,12 @@ function detect(cwd) {
     const docsUrl = urls?.get("Documentation") ?? urls?.get("documentation") ?? urls?.get("Homepage") ?? urls?.get("homepage");
     return { ecosystem: "python", name, bin, library: true, readme, docsUrl };
   }
-  const goText = readText(join6(cwd, "go.mod"));
+  const goText = readText(join7(cwd, "go.mod"));
   if (goText !== undefined) {
     const name = goText.match(/^module\s+(\S+)/m)?.[1] ?? fallback.name;
     return { ecosystem: "go", name, library: true, readme };
   }
-  const cargoText = readText(join6(cwd, "Cargo.toml"));
+  const cargoText = readText(join7(cwd, "Cargo.toml"));
   if (cargoText !== undefined) {
     const name = tomlSections(cargoText).get("package")?.get("name");
     return {
@@ -8210,10 +8500,10 @@ function renderYaml(p, tests) {
 `;
 }
 function ensureGitignore(cwd) {
-  const file = join6(cwd, ".gitignore");
+  const file = join7(cwd, ".gitignore");
   const text = readText(file);
   if (text === undefined) {
-    writeFileSync2(file, `${GITIGNORE_LINE}
+    writeFileSync3(file, `${GITIGNORE_LINE}
 `);
     return `created .gitignore with ${GITIGNORE_LINE}`;
   }
@@ -8223,12 +8513,12 @@ function ensureGitignore(cwd) {
   const sep = text === "" || text.endsWith(`
 `) ? "" : `
 `;
-  writeFileSync2(file, `${text}${sep}${GITIGNORE_LINE}
+  writeFileSync3(file, `${text}${sep}${GITIGNORE_LINE}
 `);
   return `added ${GITIGNORE_LINE} to .gitignore`;
 }
 function installSkills(cwd) {
-  const r = spawnSync3("npx", ["-y", ...SKILLS_ARGV], {
+  const r = spawnSync4("npx", ["-y", ...SKILLS_ARGV], {
     cwd,
     stdio: "inherit"
   });
@@ -8238,6 +8528,69 @@ function installSkills(cwd) {
     return `exit ${r.status ?? "by signal"}`;
   return;
 }
+async function ask(question) {
+  const rl = createInterface2({ input: process.stdin, output: process.stdout });
+  try {
+    return (await rl.question(question)).trim();
+  } finally {
+    rl.close();
+  }
+}
+async function readSecret(question) {
+  if (!process.stdin.isTTY) {
+    const chunks = [];
+    for await (const c of process.stdin)
+      chunks.push(c);
+    return Buffer.concat(chunks).toString("utf8").trim();
+  }
+  process.stdout.write(question);
+  const silent = new Writable({ write: (_c, _e, cb) => cb() });
+  const rl = createInterface2({
+    input: process.stdin,
+    output: silent,
+    terminal: true
+  });
+  try {
+    return (await rl.question("")).trim();
+  } finally {
+    rl.close();
+    process.stdout.write(`
+`);
+  }
+}
+async function chooseAuth() {
+  let current;
+  try {
+    current = resolveAuth({}).label;
+  } catch (e) {
+    current = e.message;
+  }
+  console.log(`How should the agent sign in? (now: ${current})
+  1) Login with Claude (your Claude plan)
+  2) Paste an Anthropic API key`);
+  for (;; ) {
+    const a = await ask("choose [1/2]: ");
+    if (a === "1")
+      return "claude";
+    if (a === "2")
+      return "api-key";
+    console.log("type 1 or 2");
+  }
+}
+async function setUpAuth(method) {
+  if (method === "claude") {
+    const code = claudeLoginCommand([]);
+    return code === 0 ? undefined : code;
+  }
+  const key = await readSecret("Anthropic API key (input hidden): ");
+  if (!looksLikeAnthropicKey(key)) {
+    console.error("firegoat init: that does not look like an Anthropic API key (sk-ant-...); nothing stored");
+    return EXIT.usage;
+  }
+  const file = saveApiKey(key);
+  console.log(`stored the API key in ${file} (mode 0600, not in the repo); evals pass it to the agent as ANTHROPIC_API_KEY`);
+  return;
+}
 async function askYesNo(question) {
   const rl = createInterface2({ input: process.stdin, output: process.stdout });
   try {
@@ -8261,7 +8614,18 @@ async function initCommand(args, cwd) {
   let yes = false;
   let force = false;
   let skills = false;
-  for (const a of args) {
+  let auth;
+  for (let i = 0;i < args.length; i++) {
+    const a = args[i];
+    if (a === "--auth" || a.startsWith("--auth=")) {
+      const v = a === "--auth" ? args[++i] : a.slice(7);
+      if (v !== "claude" && v !== "api-key") {
+        console.error(`firegoat init: --auth must be claude or api-key, got ${v ?? "nothing"}`);
+        return EXIT.usage;
+      }
+      auth = v;
+      continue;
+    }
     switch (a) {
       case "--yes":
       case "-y":
@@ -8276,18 +8640,18 @@ async function initCommand(args, cwd) {
         break;
       case "--help":
       case "-h":
-        console.log(INIT_USAGE);
+        console.log(INIT_HELP);
         return 0;
       default:
         console.error(`firegoat init: unknown argument: ${a}
 
-${INIT_USAGE}`);
+${INIT_HELP}`);
         return EXIT.usage;
     }
   }
-  const configPath = join6(cwd, CONFIG_FILE);
-  const keepConfig = existsSync4(configPath) && !force;
-  if (keepConfig && !skills) {
+  const configPath = join7(cwd, CONFIG_FILE);
+  const keepConfig = existsSync5(configPath) && !force;
+  if (keepConfig && !skills && !auth) {
     console.error(`firegoat init: ${CONFIG_FILE} already exists; not overwriting. Use --force to replace it.`);
     return EXIT.usage;
   }
@@ -8297,11 +8661,18 @@ ${INIT_USAGE}`);
     const project = detect(cwd);
     const tests = templates(project);
     console.log(`detected: ${describe(project)}`);
-    writeFileSync2(configPath, renderYaml(project, tests));
+    writeFileSync3(configPath, renderYaml(project, tests));
     console.log(`wrote ${CONFIG_FILE} (${tests.length} test${tests.length === 1 ? "" : "s"}: ${tests.map((t) => t.name).join(", ")}); edit the prompts to fit your users`);
   }
   console.log(ensureGitignore(cwd));
   const interactive = !yes && process.stdin.isTTY && process.stdout.isTTY;
+  if (!auth && interactive)
+    auth = await chooseAuth();
+  if (auth) {
+    const failed = await setUpAuth(auth);
+    if (failed !== undefined)
+      return failed;
+  }
   if (!skills && interactive)
     skills = await askYesNo(`Add the ${SKILLS.join(" and ")} skills to .claude/skills/ (help your agent draft better evals)?`);
   if (skills) {
@@ -8318,47 +8689,10 @@ ${INIT_USAGE}`);
   return 0;
 }
 
-// src/login.ts
-import { spawnSync as spawnSync4 } from "node:child_process";
-import { mkdirSync as mkdirSync4 } from "node:fs";
-function loginCommand(args) {
-  let claudeBin = process.env.FIREGOAT_CLAUDE_BIN || "claude";
-  for (let i = 0;i < args.length; i++) {
-    const a = args[i];
-    if (a === "--claude-bin" && args[i + 1] !== undefined)
-      claudeBin = args[++i];
-    else if (a.startsWith("--claude-bin="))
-      claudeBin = a.slice(13);
-    else {
-      console.error(`firegoat login: unknown argument: ${a}`);
-      return EXIT.usage;
-    }
-  }
-  const configDir = claudeConfigDir();
-  mkdirSync4(configDir, { recursive: true });
-  const env = {
-    ...process.env,
-    CLAUDE_CONFIG_DIR: configDir
-  };
-  delete env.ANTHROPIC_API_KEY;
-  console.error(`Signing claude in to ${configDir}. Evals run with this login, never your own ~/.claude.`);
-  const r = spawnSync4(claudeBin, ["auth", "login"], { env, stdio: "inherit" });
-  if (r.error) {
-    console.error(`firegoat login: could not run ${claudeBin}: ${r.error.message}`);
-    return EXIT.notGraded;
-  }
-  if (r.status !== 0) {
-    console.error(`firegoat login: claude auth login exited ${r.status}`);
-    return EXIT.notGraded;
-  }
-  console.log("logged in. next: npx firegoat eval");
-  return EXIT.pass;
-}
-
 // src/open.ts
 import { spawnSync as spawnSync5 } from "node:child_process";
-import { existsSync as existsSync5, readdirSync } from "node:fs";
-import { join as join7 } from "node:path";
+import { existsSync as existsSync6, readdirSync } from "node:fs";
+import { join as join8 } from "node:path";
 var OPENER = {
   darwin: "open",
   linux: "xdg-open"
@@ -8367,6 +8701,10 @@ function openCommand(args, cwd) {
   let print = false;
   const ids = [];
   for (const a of args) {
+    if (a === "--help" || a === "-h") {
+      console.log(OPEN_HELP);
+      return EXIT.pass;
+    }
     if (a === "--print")
       print = true;
     else if (a.startsWith("-")) {
@@ -8379,8 +8717,8 @@ function openCommand(args, cwd) {
     console.error("firegoat open: takes at most one run id");
     return EXIT.usage;
   }
-  const base = join7(findConfigDir(cwd), ".firegoat", "runs");
-  const runs = existsSync5(base) ? readdirSync(base).sort() : [];
+  const base = join8(findConfigDir(cwd), ".firegoat", "runs");
+  const runs = existsSync6(base) ? readdirSync(base).sort() : [];
   const id = ids[0] ?? runs.at(-1);
   if (!id) {
     console.error(`firegoat open: no runs in ${base}. Run: npx firegoat eval`);
@@ -8390,7 +8728,7 @@ function openCommand(args, cwd) {
     console.error(`firegoat open: no run ${JSON.stringify(id)} in ${base} (latest: ${runs.at(-1) ?? "none"})`);
     return EXIT.usage;
   }
-  const dir = join7(base, id);
+  const dir = join8(base, id);
   console.log(dir);
   const opener = OPENER[process.platform];
   if (print || !opener)
@@ -8404,32 +8742,6 @@ function openCommand(args, cwd) {
 }
 
 // src/cli.ts
-var USAGE = `firegoat ${package_default.version}: local-first AX evals
-
-Usage:
-  firegoat init [--yes] [--force] [--skills]   write firegoat.yaml with starter evals
-  firegoat login [--claude-bin PATH]          sign claude in to firegoat's own config dir (once)
-  firegoat eval [test..] [--runs N] [--var KEY=value].. [--json] [--host] [--claude-bin PATH]
-                                               run evals with claude in a sandboxed fresh workdir
-  firegoat open [runId] [--print]              open a run's folder (default: the latest)
-  firegoat --version
-
-eval options:
-  test..             only run these tests (by name)
-  --runs N           runs per test (overrides runs: in firegoat.yaml)
-  --var KEY=value    fills {{vars.KEY}} in prompts and checks; env for setup, workdir_setup and checks
-  --json             JSON summary on stdout, progress on stderr
-  --host             no sandbox (needed off macOS; use a throwaway container)
-  --claude-bin PATH  claude binary to run (env: FIREGOAT_CLAUDE_BIN; default: claude on PATH)
-
-firegoat.yaml is looked up in the current dir, then parents up to the git root.
-
-Exit codes:
-  0  every run passed
-  1  a goal or check failed
-  2  at least one run was not graded (claude missing, auth, crash, timeout); wins over 1
-  3  usage or config error (bad flag, unknown test, invalid or missing firegoat.yaml)
-`;
 async function main(argv) {
   const [cmd, ...rest] = argv;
   switch (cmd) {
@@ -8440,12 +8752,24 @@ async function main(argv) {
     case undefined:
     case "--help":
     case "-h":
-      console.log(USAGE);
+      console.log(topHelp(package_default.version));
       return 0;
     case "init":
       return initCommand(rest, process.cwd());
+    case "claude":
+      if (rest[0] === "login")
+        return claudeLoginCommand(rest.slice(1));
+      if (rest[0] === "--help" || rest[0] === "-h") {
+        console.log(CLAUDE_LOGIN_HELP);
+        return 0;
+      }
+      console.error(`firegoat claude: unknown subcommand
+
+${CLAUDE_LOGIN_HELP}`);
+      return EXIT.usage;
     case "login":
-      return loginCommand(rest);
+      console.error("firegoat login is deprecated for the Claude sign-in: use `firegoat claude login`");
+      return claudeLoginCommand(rest);
     case "eval":
       return evalCommand(rest, process.cwd());
     case "open":
@@ -8453,7 +8777,7 @@ async function main(argv) {
     default:
       console.error(`unknown command: ${cmd}
 
-${USAGE}`);
+${topHelp(package_default.version)}`);
       return EXIT.usage;
   }
 }