@bjoernrochel/sevro 1.0.0 → 1.1.0

package/docs/recipes.md

@@ -7,14 +7,15 @@ small extension.
 
 ## Prepare a recipe project
 
-Use Bun 1.3.13, Git, Node 24.13.0, and `rg`. Shell checks need macOS
+Use Node.js 24.13.0 or newer, Git, and `rg`. Bun is optional for the installed
+recipes; the same JavaScript files also run with Bun 1.3.13. Shell checks need macOS
 `sandbox-exec` or Ubuntu `bubblewrap`; native Claude also needs `socat` on
 Ubuntu. Python/UV and mise add the prerequisites stated below.
 [Install Sevro](installing.md), then set the recipe path from the directory
 containing its local installation:
 
 ```sh
-export SEVRO_RECIPES="$PWD/node_modules/@bjoernrochel/sevro/examples/recipes"
+export SEVRO_RECIPES="$PWD/node_modules/@bjoernrochel/sevro/dist/examples/recipes"
 ```
 
 Run commands below from an otherwise empty working directory to keep generated
@@ -24,7 +25,7 @@ the CLI beside its original package; it does not require a global installation.
 To adapt a separate copy, preserve the small project's whole directory:
 
 ```sh
-export SEVRO_RECIPE_CLI="$(cd "$SEVRO_RECIPES/../../src" && pwd)/cli.ts"
+export SEVRO_RECIPE_CLI="$(cd "$SEVRO_RECIPES/../../src" && pwd)/cli.js"
 cp -R "$SEVRO_RECIPES" ./my-recipes
 export SEVRO_RECIPES="$PWD/my-recipes"
 ```
@@ -45,7 +46,7 @@ deliberately exits `1` with execution/grading `completed` and task `failed`.
 Run negative controls separately when using a shell with `set -e`. A missing
 tool is a prerequisite error, not a passing or intentionally failing control.
 
-`candidate.ts` is an explicitly trusted deterministic fixture editor. It changes
+`candidate.js` (built from `candidate.ts`) is an explicitly trusted deterministic fixture editor. It changes
 the arithmetic implementation for the coding recipe and supplies a fixed answer
 for the other controls. It does not simulate native tool permissions: the real
 isolated shell graders execute the stated tools and assertions. Native routes
@@ -59,16 +60,16 @@ available identifier and supply your private authentication file path:
 ```sh
 export SEVRO_RECIPE_MODEL='replace-with-an-available-Codex-model'
 export SEVRO_RECIPE_CODEX_AUTH="${CODEX_HOME:-$HOME/.codex}/auth.json"
-bun "$SEVRO_RECIPES/run.ts" coding-change codex --dry
-bun "$SEVRO_RECIPES/run.ts" coding-change codex
+node "$SEVRO_RECIPES/run.js" coding-change codex --dry
+node "$SEVRO_RECIPES/run.js" coding-change codex
 ```
 
 For Claude, use its supported credential route and a Claude model identifier:
 
 ```sh
 export SEVRO_RECIPE_MODEL='replace-with-an-available-Claude-model'
-bun "$SEVRO_RECIPES/run.ts" coding-change claude --dry
-bun "$SEVRO_RECIPES/run.ts" coding-change claude
+node "$SEVRO_RECIPES/run.js" coding-change claude --dry
+node "$SEVRO_RECIPES/run.js" coding-change claude
 ```
 
 Substitute the other recipe name to run its native task. `--dry` prepares and
@@ -86,13 +87,13 @@ Files: [coding-change case](../examples/recipes/cases/coding-change.json),
 The fixture contains a broken `math.cjs` and a task README. The task asks the
 participant to implement addition. The evaluator checks positive, negative and
 zero arguments. Its source stays outside the candidate fixture: the launcher
-transpiles the checked TypeScript test and embeds its JavaScript in the
+reads the compiled evaluator test and embeds its JavaScript in the
 evaluator-owned shell command before execution.
 
 <!-- sevro-example:recipe-coding-change-pass -->
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" coding-change deterministic
+node "$SEVRO_RECIPES/run.js" coding-change deterministic
 ```
 
 The isolated check first asserts that the original evaluator test is unreadable,
@@ -103,7 +104,7 @@ in place, so the arithmetic assertion fails:
 <!-- sevro-example:recipe-coding-change-fail -->
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" coding-change deterministic --fail
+node "$SEVRO_RECIPES/run.js" coding-change deterministic --fail
 ```
 
 Use a native route below to evaluate an actual coding agent. Adding a participant
@@ -150,7 +151,7 @@ process must find exactly that line.
 <!-- sevro-example:recipe-host-tools-pass -->
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" host-tools deterministic
+node "$SEVRO_RECIPES/run.js" host-tools deterministic
 ```
 
 The negative fixture contains port `9000`, so that same search fails:
@@ -158,7 +159,7 @@ The negative fixture contains port `9000`, so that same search fails:
 <!-- sevro-example:recipe-host-tools-fail -->
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" host-tools deterministic --fail
+node "$SEVRO_RECIPES/run.js" host-tools deterministic --fail
 ```
 
 ### Node
@@ -170,7 +171,7 @@ Node process parses it, asserts the port, and prints `3000`.
 <!-- sevro-example:recipe-node-pass -->
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" node deterministic
+node "$SEVRO_RECIPES/run.js" node deterministic
 ```
 
 The negative fixture sets port `9000`, causing the Node assertion to fail:
@@ -178,7 +179,7 @@ The negative fixture sets port `9000`, causing the Node assertion to fail:
 <!-- sevro-example:recipe-node-fail -->
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" node deterministic --fail
+node "$SEVRO_RECIPES/run.js" node deterministic --fail
 ```
 
 ### Python with UV
@@ -198,7 +199,7 @@ port and prints `3000`:
 <!-- sevro-example:recipe-python-uv-pass -->
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" python-uv deterministic
+node "$SEVRO_RECIPES/run.js" python-uv deterministic
 ```
 
 The negative fixture changes the port to `9000`; Python raises an assertion:
@@ -206,7 +207,7 @@ The negative fixture changes the port to `9000`; Python raises an assertion:
 <!-- sevro-example:recipe-python-uv-fail -->
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" python-uv deterministic --fail
+node "$SEVRO_RECIPES/run.js" python-uv deterministic --fail
 ```
 
 For real dependencies, prepare a public cache once outside the evaluated project
@@ -238,7 +239,7 @@ sandbox or grant mise's whole data directory.
 <!-- sevro-example:recipe-mise-pass -->
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" mise deterministic
+node "$SEVRO_RECIPES/run.js" mise deterministic
 ```
 
 The negative fixture changes the port to `9000`, failing the same Node check:
@@ -246,7 +247,7 @@ The negative fixture changes the port to `9000`, failing the same Node check:
 <!-- sevro-example:recipe-mise-fail -->
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" mise deterministic --fail
+node "$SEVRO_RECIPES/run.js" mise deterministic --fail
 ```
 
 An attempted isolated `mise exec` with read-only operator data returned
@@ -274,13 +275,13 @@ and retains the artifacts. Expect exit `0`, execution `not_run`, grading
 <!-- sevro-example:recipe-repository-skill-codex-dry -->
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" repository-skill codex --dry
+node "$SEVRO_RECIPES/run.js" repository-skill codex --dry
 ```
 
 <!-- sevro-example:recipe-repository-skill-claude-dry -->
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" repository-skill claude --dry
+node "$SEVRO_RECIPES/run.js" repository-skill claude --dry
 ```
 
 For a real Codex trial, export `SEVRO_RECIPE_MODEL` to an available Codex model
@@ -291,15 +292,15 @@ remove `--dry`. Credentials stay outside recipe/project inputs and argv.
 These calls consume account quota.
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" repository-skill codex
-bun "$SEVRO_RECIPES/run.ts" repository-skill codex --fail
+node "$SEVRO_RECIPES/run.js" repository-skill codex
+node "$SEVRO_RECIPES/run.js" repository-skill codex --fail
 ```
 
 With `SEVRO_RECIPE_MODEL` set for Claude instead:
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" repository-skill claude
-bun "$SEVRO_RECIPES/run.ts" repository-skill claude --fail
+node "$SEVRO_RECIPES/run.js" repository-skill claude
+node "$SEVRO_RECIPES/run.js" repository-skill claude --fail
 ```
 
 The neutral invocation placeholder becomes `$startup-guide` for Codex and
@@ -327,13 +328,13 @@ construction, verdict parsing, and retained semantic evidence:
 <!-- sevro-example:recipe-semantic-pass -->
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" semantic deterministic
+node "$SEVRO_RECIPES/run.js" semantic deterministic
 ```
 
 <!-- sevro-example:recipe-semantic-fail -->
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" semantic deterministic --fail
+node "$SEVRO_RECIPES/run.js" semantic deterministic --fail
 ```
 
 The protocol judge derives pass/fail from the actual candidate-response field
@@ -345,8 +346,8 @@ For an actual model candidate and judge, set the native variables above and
 export `SEVRO_RECIPE_JUDGE_MODEL` to an available Codex judge model:
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" semantic codex --dry
-bun "$SEVRO_RECIPES/run.ts" semantic codex
+node "$SEVRO_RECIPES/run.js" semantic codex --dry
+node "$SEVRO_RECIPES/run.js" semantic codex
 ```
 
 `semantic claude` selects a Claude candidate with a separate Codex judge;
@@ -367,12 +368,12 @@ a grading error, not a failed proposition or a passing task. Read
 Files: [extension.ts](../examples/recipes/extension.ts),
 [methods](../examples/recipes/extension-methods.ts), and
 [artifact preparation](../examples/recipes/extension-artifacts.ts). Prerequisites:
-Bun, Git and the platform's shell isolation tool. No model is required:
+Node or Bun, Git and the platform's shell isolation tool. No model is required:
 
 <!-- sevro-example:recipe-lifecycle-pass -->
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" lifecycle deterministic
+node "$SEVRO_RECIPES/run.js" lifecycle deterministic
 ```
 
 `describe` advertises version/capabilities/grader; `resolve` returns the task
@@ -387,7 +388,7 @@ response; diagnostics belong on stderr.
 <!-- sevro-example:recipe-lifecycle-fail -->
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" lifecycle deterministic --fail
+node "$SEVRO_RECIPES/run.js" lifecycle deterministic --fail
 ```
 
 The wrong final answer fails the built-in output assertion and extension check,
@@ -413,13 +414,13 @@ First add only native goals to the empty runtime and validate preparation:
 <!-- sevro-example:recipe-goals-codex-dry -->
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" plugins codex --dry --goals-only
+node "$SEVRO_RECIPES/run.js" plugins codex --dry --goals-only
 ```
 
 <!-- sevro-example:recipe-goals-claude-dry -->
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" plugins claude --dry --goals-only
+node "$SEVRO_RECIPES/run.js" plugins claude --dry --goals-only
 ```
 
 Then add `hooks.plugins: ["startup"]` so only this prepared plugin's command
@@ -429,13 +430,13 @@ Claude uses plugin-only settings, so do not add `--claude-project-settings`:
 <!-- sevro-example:recipe-plugins-codex-dry -->
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" plugins codex --dry
+node "$SEVRO_RECIPES/run.js" plugins codex --dry
 ```
 
 <!-- sevro-example:recipe-plugins-claude-dry -->
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" plugins claude --dry
+node "$SEVRO_RECIPES/run.js" plugins claude --dry
 ```
 
 All four dry commands expect `not_run`/`not_requested`/`not_assessed`, exit `0`.
@@ -454,15 +455,15 @@ with network denied and only the prepared plugin subtree readable.
 facility or passing this startup task alone does not prove goal completion.
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" plugins codex
-bun "$SEVRO_RECIPES/run.ts" plugins codex --fail
+node "$SEVRO_RECIPES/run.js" plugins codex
+node "$SEVRO_RECIPES/run.js" plugins codex --fail
 ```
 
 Or, with the Claude model/credential route configured:
 
 ```sh
-bun "$SEVRO_RECIPES/run.ts" plugins claude
-bun "$SEVRO_RECIPES/run.ts" plugins claude --fail
+node "$SEVRO_RECIPES/run.js" plugins claude
+node "$SEVRO_RECIPES/run.js" plugins claude --fail
 ```
 
 For a deliberate failure, run that live command with `--fail`; the requested

package/docs/releases.md

@@ -12,7 +12,21 @@ Published `rc.1` and `rc.2` retain their original BSL grants. Always select a ne
 version for publication; a local development tarball does not replace an
 already published artifact.
 
-## Unreleased: smaller npm package
+<a id="unreleased-node-runtime-support"></a>
+
+## Node runtime support
+
+The package ships compiled JavaScript for Node.js 24.13.0 or newer and Bun
+1.3.13. Development retains Bun. Builds populate `dist/`; the package phase
+checks both runtimes and includes Node workflows with Bun absent from PATH.
+`runnerBuildDigest` includes compiled output for installed execution, so this
+changes build identities. Existing evidence keeps its original identity.
+Release preparation builds first, then packs with lifecycle scripts disabled;
+this adds no release test rerun.
+
+<a id="unreleased-smaller-npm-package"></a>
+
+## Package contents and size
 
 The npm package omits logo images, historical validation and usability pages,
 and raw evidence, including the Darrow migration patch and old guide verification
@@ -20,10 +34,11 @@ record. All remain in the repository; packaged pages link to their online copies
 The current `LICENSE`, contribution terms, branding permissions and retained
 earlier-release license text still ship. No licensing or branding grants change.
 
-For this cleanup, `npm pack --dry-run --ignore-scripts --json` measured the
+During the package-size cleanup, `npm pack --dry-run --ignore-scripts --json` measured a
 development tarball at about 0.39 MB compressed, down from 3,545,849 bytes, and
 about 1.86 MB unpacked, down from 6,632,891 bytes (about 89% and 72% smaller).
-These are candidate measurements, not a newly published release.
+These historical development-build measurements do not guarantee the size of
+every release.
 
 `runnerBuildDigest` hashes packaged paths and bytes under `docs/` as well as the
 runtime and other package inputs. This packaging change therefore causes a
@@ -42,7 +57,8 @@ including a release candidate. Use `next` for a preview that should require an
 explicit opt-in. Preparation accepts either tag for prerelease and stable versions.
 
 Preparation creates a new absolute output directory and never replaces an
-existing directory. It packs with lifecycle scripts disabled, retains the npm
+existing directory. `release:prepare` first builds the JavaScript artifact,
+then packs with lifecycle scripts disabled, retains the npm
 file inventory, and writes `release.json` and `SHA256SUMS` beside the tarball.
 The record identifies the package, version, release tag, distribution tag,
 tarball SHA-256 and npm SHA-512 integrity, and the preparation runtimes.

package/docs/run-settings-v1.md

@@ -7,14 +7,14 @@ Save recurring evaluation selections in a JSON run manifest, then use
 ## Repeat an evaluation
 
 Start with an existing [resolved case and adapter](getting-started.md), and save
-this `evaluation.json` beside `case.json` and `host-adapter.ts`:
+this `evaluation.json` beside `case.json` and `host-adapter.js`:
 
 ```json
 {
   "format": "sevro.run.v1",
   "settings": {
     "case-file": "./case.json",
-    "adapter-module": "./host-adapter.ts",
+    "adapter-module": "./host-adapter.js",
     "project-root": ".",
     "condition": "passive",
     "trials": 3,
@@ -95,7 +95,7 @@ occurrences build the replacement array. `--no-dry`, `--no-shell-isolation`,
 `true` values with `false`.
 
 Selecting an alternative route on the CLI removes its saved alternative and
-route-only settings. For example, `--adapter-module ./host-adapter.ts` replaces
+route-only settings. For example, `--adapter-module ./host-adapter.js` replaces
 a saved native candidate and its model/effort. Selecting `--host claude` over
 a saved Codex candidate requires fresh explicit `--model` and `--effort`;
 grading roles keep their separate selections. Shared Codex settings remain

package/docs/specs/repository-guide.md

@@ -1,6 +1,6 @@
 # Repository guide
 
-The repository-scoped `sevro-guide` explains this checkout. It is independent
+The repository-scoped `sevro-guide` explains Sevro from repository sources. It is independent
 of npm packaging and serves a static documentation spine, not an execution agent.
 
 ## Contract

package/docs/troubleshooting.md

@@ -5,7 +5,7 @@ cause is not a diagnosis of your machine.
 
 | Symptom                                   | Check and next step                                                                                                                                                                     |
 | ----------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
-| Command cannot run                        | Use Bun 1.3.13 and verify the local package/version. Read [installation](installing.md).                                                                                                |
+| Command cannot run                        | Use Node.js 24.13.0 or newer, or the explicit Bun 1.3.13 invocation, and verify the local package/version. Read [installation](installing.md).                                          |
 | Project snapshot contains a symbolic link | Use a separate evaluated project directory as in [the tutorial](getting-started.md); do not include installed dependencies in a plain-directory snapshot.                               |
 | Exit 0 but no passing task                | Prompt-only/dry runs are `not_assessed`. Read [results](reading-results.md).                                                                                                            |
 | Exit 64 before execution                  | Check absolute paths, exclusive route options, configuration, and capabilities in the [CLI reference](development-cli.md).                                                              |

package/docs/typescript-quality.md

@@ -25,7 +25,7 @@ order; `--fast` runs only `static` without installing dependencies.
 | `static`  | Schema freshness, documentation contracts, formatting, type-aware lint and typing             |
 | `docs`    | All guide dry runs, tutorial examples and recipe checks                                       |
 | `tests`   | Coverage integrity controls, the complete instrumented Bun suite and both coverage thresholds |
-| `package` | Packed-package installation and installed CLI workflows                                       |
+| `package` | Packed-package installation and installed CLI workflows on Bun and Node                       |
 
 Use the same prerequisites as the corresponding complete-gate checks. A single
 phase is partial evidence; full acceptance requires all four on the same candidate.
@@ -46,6 +46,9 @@ need no manual manifest edit. Non-ignored untracked sources are included in loca
 candidate checks; committed CI checkouts use their tracked paths. Declarations
 take precedence over directory roles, and only the validators in the shared
 [schema catalog](../scripts/schema-catalog.ts) receive the generated disposition.
+Generated `dist/` output is ignored by Git, lint and the authored-source inventory.
+It is rebuilt from inventoried TypeScript and validated through installed-package
+workflows; it never substitutes for authored production coverage.
 Unknown or ambiguous paths, missing tracked files, an incomplete generated set,
 and sources outside lint or typing are refused. An independent filesystem audit
 also refuses ignored authored sources rather than silently dropping them.
@@ -203,6 +206,20 @@ deliberately exit 1. A five-of-five statement, one-of-two branch sample must fai
 with `Below 95%: branches`. The integration suite retains real CLI/extension,
 graceful SIGINT/SIGTERM cancellation, filesystem isolation and owned-child paths.
 
+## Installed runtime acceptance
+
+Bun 1.3.13 remains the contributor runtime and test runner. `bun run build`
+emits JavaScript and source maps under `dist/`; `prepack` builds before ordinary
+`npm pack`. Packing with lifecycle scripts disabled requires an explicit build.
+`bun run test:runtime-node` installs the candidate archive with npm, removes Bun
+from the consumer PATH, and exercises the Node 24.13.0 CLI, templates and recipes.
+It also interrupts a real isolated shell check, verifies the interruption exit
+and retained evidence, and refuses a surviving owned child process.
+It reuses the installed-package workflow checks, including synthetic native
+fixtures, with Node selected explicitly. The package quality phase requires
+both the existing Bun package checks and Node acceptance. These deterministic
+checks do not establish authenticated model behavior.
+
 ## Environment and CI
 
 The discovery backend was observed with Bun 1.3.13, macOS arm64 and Node 24.13.0.

package/examples/recipes/executable.ts

@@ -0,0 +1,19 @@
+import { accessSync, constants, statSync } from "node:fs";
+import { delimiter, resolve } from "node:path";
+
+export function which(name: string, path?: string): string | null {
+  for (const directory of executablePath(path).split(delimiter)) {
+    const candidate = resolve(directory, name);
+    try {
+      accessSync(candidate, constants.X_OK);
+      if (statSync(candidate).isFile()) return candidate;
+    } catch {
+      /* Continue searching the declared executable path. */
+    }
+  }
+  return null;
+}
+
+function executablePath(path: string | undefined): string {
+  return path ?? process.env.PATH ?? "";
+}

package/examples/recipes/extension-artifacts.ts

@@ -16,7 +16,7 @@ export async function repositoryArtifacts(host: string) {
   const directory = host === "claude" ? ".claude" : ".agents";
   const relative = `${directory}/skills/startup-guide/SKILL.md`;
   const body = await readFile(
-    join(import.meta.dir, "repository-skill", relative),
+    join(import.meta.dirname, "repository-skill", relative),
     "utf8",
   );
   return [artifact(relative, body)];

package/examples/recipes/extension.ts

@@ -1,3 +1,4 @@
+import { readFileSync } from "node:fs";
 import {
   describe,
   evaluate,
@@ -9,7 +10,7 @@ import {
 const mode = process.argv[2] ?? "lifecycle";
 const host = process.argv[3] ?? "deterministic";
 const fail = process.argv[4] === "fail";
-const request = record(JSON.parse(await Bun.stdin.text()) as unknown);
+const request = record(JSON.parse(readFileSync(0, "utf8")) as unknown);
 
 async function result() {
   switch (request.method) {

package/examples/recipes/modules.ts

@@ -0,0 +1,9 @@
+import { join } from "node:path";
+
+/** Locate source modules for Bun development or compiled modules for consumers. */
+export function recipeFile(...parts: string[]): string {
+  const path = join(import.meta.dirname, ...parts);
+  return import.meta.filename.endsWith(".ts")
+    ? path
+    : path.replace(/\.ts$/, ".js");
+}

package/examples/recipes/run-inputs.ts

@@ -1,3 +1,4 @@
+import { recipeFile } from "./modules";
 import { readFile, writeFile } from "node:fs/promises";
 import { join } from "node:path";
 import { record } from "./extension-methods";
@@ -7,6 +8,11 @@ function shellQuote(text: string): string {
   return `'${text.replaceAll("'", "'\\''")}'`;
 }
 
+function evaluatorCode(test: string): string {
+  if (import.meta.filename.endsWith(".js")) return test;
+  return new Bun.Transpiler({ loader: "ts" }).transformSync(test);
+}
+
 async function caseFile(
   options: RecipeOptions,
   state: string,
@@ -14,22 +20,17 @@ async function caseFile(
   const value = record(
     JSON.parse(
       await readFile(
-        join(import.meta.dir, "cases", `${options.recipe}.json`),
+        join(import.meta.dirname, "cases", `${options.recipe}.json`),
         "utf8",
       ),
     ) as unknown,
   );
   if (options.recipe === "coding-change") {
-    const test = await readFile(
-      join(import.meta.dir, "coding-change/check.ts"),
-      "utf8",
-    );
+    const test = await readFile(recipeFile("coding-change/check.ts"), "utf8");
     const checks: unknown[] = Array.isArray(value.checks) ? value.checks : [];
     const check = record(checks[0]);
-    const javascript = new Bun.Transpiler({ loader: "ts" }).transformSync(test);
-    const protectedTest = shellQuote(
-      join(import.meta.dir, "coding-change/check.ts"),
-    );
+    const javascript = evaluatorCode(test);
+    const protectedTest = shellQuote(recipeFile("coding-change/check.ts"));
     record(check.configuration).run =
       `test ! -r ${protectedTest}\nnode --input-type=module -e ${shellQuote(javascript)}`;
   }
@@ -62,7 +63,7 @@ async function extensionFiles(
     command,
     JSON.stringify([
       process.execPath,
-      join(import.meta.dir, "extension.ts"),
+      recipeFile("extension.ts"),
       options.recipe,
       options.route,
       options.fail ? "fail" : "pass",
@@ -80,10 +81,7 @@ async function extensionFiles(
     command,
     "--case-id",
     "recipe-extension",
-    ...sources.flatMap((path) => [
-      "--extension-source-file",
-      join(import.meta.dir, path),
-    ]),
+    ...sources.flatMap((path) => ["--extension-source-file", recipeFile(path)]),
   ];
 }
 
@@ -118,7 +116,7 @@ export async function hostArguments(
   state: string,
 ): Promise<string[]> {
   if (options.route === "deterministic")
-    return ["--adapter-module", join(import.meta.dir, "candidate.ts")];
+    return ["--adapter-module", recipeFile("candidate.ts")];
   const model = selectedModel(options.dry, "SEVRO_RECIPE_MODEL");
   if (!model)
     throw new Error(
@@ -156,10 +154,7 @@ export async function semanticArguments(
 ): Promise<string[]> {
   if (options.recipe !== "semantic") return [];
   if (options.route === "deterministic")
-    return [
-      "--semantic-adapter-module",
-      join(import.meta.dir, "semantic-judge.ts"),
-    ];
+    return ["--semantic-adapter-module", recipeFile("semantic-judge.ts")];
   const model = selectedModel(options.dry, "SEVRO_RECIPE_JUDGE_MODEL");
   if (!model)
     throw new Error(

package/examples/recipes/run-options.ts

@@ -1,3 +1,5 @@
+import { spawnSync } from "node:child_process";
+import { which } from "./executable";
 import { delimiter, isAbsolute, join } from "node:path";
 import { homedir } from "node:os";
 
@@ -60,7 +62,7 @@ function validateGoalsStep(flags: string[]): void {
 }
 
 export function requireTool(name: string): string {
-  const path = Bun.which(name);
+  const path = which(name);
   if (!path)
     throw new Error(
       `Recipe prerequisite missing: install ${name} and put it on PATH`,
@@ -78,7 +80,7 @@ export function runtimeFile(options: RecipeOptions): string {
     plugins: options.goalsOnly ? "goals" : "plugins",
   };
   return join(
-    import.meta.dir,
+    import.meta.dirname,
     "runtime",
     `${files[options.recipe] ?? "empty"}.json`,
   );
@@ -99,19 +101,18 @@ function requireStandaloneRipgrep(): void {
   const paths = (process.env.PATH ?? "")
     .split(delimiter)
     .filter((entry) => !roots.some((root) => entry.startsWith(root)));
-  if (!Bun.which("rg", { PATH: paths.join(delimiter) }))
+  if (!which("rg", paths.join(delimiter)))
     throw new Error(
       "Install standalone ripgrep outside Codex/Claude credential directories, then put it on PATH (for example brew install ripgrep or apt-get install ripgrep)",
     );
 }
 
 function selectMiseNode(): void {
-  const result = Bun.spawnSync([requireTool("mise"), "where", "node@24.13.0"], {
-    stdout: "pipe",
-    stderr: "pipe",
+  const result = spawnSync(requireTool("mise"), ["where", "node@24.13.0"], {
+    encoding: "utf8",
   });
-  const installation = result.stdout.toString().trim();
-  if (result.exitCode !== 0 || !isAbsolute(installation))
+  const installation = result.stdout.trim();
+  if (result.status !== 0 || !isAbsolute(installation))
     throw new Error(
       "Install the recipe's Node with mise install node@24.13.0 before running it",
     );

package/examples/recipes/run.ts

@@ -1,3 +1,5 @@
+import { spawn } from "node:child_process";
+import { Readable } from "node:stream";
 import { mkdir, realpath, writeFile } from "node:fs/promises";
 import { join, resolve } from "node:path";
 import { hostArguments, recipeInputs, semanticArguments } from "./run-inputs";
@@ -20,7 +22,13 @@ await mkdir(join(state, "project"), { recursive: true });
 await mkdir(join(state, "config"), { recursive: true });
 const root = await realpath(state);
 const cli =
-  process.env.SEVRO_RECIPE_CLI ?? resolve(import.meta.dir, "../../src/cli.ts");
+  process.env.SEVRO_RECIPE_CLI ??
+  resolve(
+    import.meta.dirname,
+    import.meta.filename.endsWith(".ts")
+      ? "../../src/cli.ts"
+      : "../../src/cli.js",
+  );
 const args = [
   process.execPath,
   cli,
@@ -42,20 +50,26 @@ const args = [
   "1",
   "--shell-isolation",
   "--protected-root",
-  import.meta.dir,
+  import.meta.dirname,
   ...(await recipeInputs(options, root)),
   ...(await hostArguments(options, root)),
   ...(await semanticArguments(options, root)),
   ...(options.dry ? ["--dry"] : []),
 ];
-const child = Bun.spawn(args, {
+const child = spawn(process.execPath, args.slice(1), {
   env: { ...process.env, SEVRO_RECIPE_FAIL: options.fail ? "1" : "0" },
-  stdout: "pipe",
-  stderr: "inherit",
+  stdio: ["ignore", "pipe", "inherit"],
 });
 const [output, code] = await Promise.all([
-  new Response(child.stdout).text(),
-  child.exited,
+  new Response(
+    Readable.toWeb(child.stdout) as unknown as ReadableStream<Uint8Array>,
+  ).text(),
+  new Promise<number>((resolve, reject) => {
+    child.once("error", reject);
+    child.once("close", (code) => {
+      resolve(code ?? 1);
+    });
+  }),
 ]);
 await writeFile(join(root, "result.json"), output);
 process.stderr.write(`Recipe result: ${join(root, "result.json")}\n`);