@ai-hero/sandcastle 0.6.4 → 0.6.6

package/dist/InitService.js

@@ -1,733 +0,0 @@
-import { FileSystem } from "@effect/platform";
-import { Effect } from "effect";
-import { dirname, join } from "node:path";
-import { fileURLToPath } from "node:url";
-import { SANDBOX_REPO_DIR } from "./SandboxFactory.js";
-const GITIGNORE = `.env
-logs/
-worktrees/
-`;
-/**
- * Filename of the setup prompt scaffolded for the `custom` issue tracker.
- * Both the per-agent `setupCommand` and the in-scaffold sentinels point at it,
- * so it is defined once here.
- */
-const SETUP_ISSUE_TRACKER_DOC = "SETUP_ISSUE_TRACKER.md";
-const SETUP_ISSUE_TRACKER_PATH = `.sandcastle/${SETUP_ISSUE_TRACKER_DOC}`;
-const TEMPLATES = [
-    {
-        name: "blank",
-        description: "Bare scaffold — write your own prompt and orchestration",
-    },
-    {
-        name: "simple-loop",
-        description: "Picks issues one by one and closes them",
-    },
-    {
-        name: "sequential-reviewer",
-        description: "Implements issues one by one, with a code review step after each",
-    },
-    {
-        name: "parallel-planner",
-        description: "Plans parallelizable issues, executes on separate branches, merges",
-    },
-    {
-        name: "parallel-planner-with-review",
-        description: "Plans parallelizable issues, executes with per-branch review, merges",
-    },
-];
-export const listTemplates = () => TEMPLATES;
-const CLAUDE_CODE_DOCKERFILE = `FROM node:22-bookworm
-
-# Install system dependencies
-RUN apt-get update && apt-get install -y \\
-  git \\
-  curl \\
-  jq \\
-  && rm -rf /var/lib/apt/lists/*
-
-{{ISSUE_TRACKER_TOOLS}}
-
-# Build-args for UID/GID alignment: sandcastle docker build-image
-# defaults these to the host user's UID/GID so image-built files
-# and bind-mounted files share an owner without runtime chown.
-ARG AGENT_UID=1000
-ARG AGENT_GID=1000
-
-# Rename the base image's "node" user to "agent" and align UID/GID.
-RUN groupmod -o -g $AGENT_GID node && usermod -o -u $AGENT_UID -g $AGENT_GID -d /home/agent -m -l agent node
-USER \${AGENT_UID}:\${AGENT_GID}
-
-# Install Claude Code CLI
-RUN curl -fsSL https://claude.ai/install.sh | bash
-
-# Add Claude to PATH
-ENV PATH="/home/agent/.local/bin:$PATH"
-
-WORKDIR /home/agent
-
-# In worktree sandbox mode, Sandcastle bind-mounts the git worktree at ${SANDBOX_REPO_DIR}
-# and overrides the working directory to ${SANDBOX_REPO_DIR} at container start.
-# Structure your Dockerfile so that ${SANDBOX_REPO_DIR} can serve as the project root.
-ENTRYPOINT ["sleep", "infinity"]
-`;
-const PI_DOCKERFILE = `FROM node:22-bookworm
-
-# Install system dependencies
-RUN apt-get update && apt-get install -y \\
-  git \\
-  curl \\
-  jq \\
-  && rm -rf /var/lib/apt/lists/*
-
-{{ISSUE_TRACKER_TOOLS}}
-
-# Build-args for UID/GID alignment: sandcastle docker build-image
-# defaults these to the host user's UID/GID so image-built files
-# and bind-mounted files share an owner without runtime chown.
-ARG AGENT_UID=1000
-ARG AGENT_GID=1000
-
-# Rename the base image's "node" user to "agent" and align UID/GID.
-RUN groupmod -o -g $AGENT_GID node && usermod -o -u $AGENT_UID -g $AGENT_GID -d /home/agent -m -l agent node
-
-# Install pi coding agent (run as root before USER agent)
-RUN npm install -g @mariozechner/pi-coding-agent
-
-USER \${AGENT_UID}:\${AGENT_GID}
-
-WORKDIR /home/agent
-
-# In worktree sandbox mode, Sandcastle bind-mounts the git worktree at ${SANDBOX_REPO_DIR}
-# and overrides the working directory to ${SANDBOX_REPO_DIR} at container start.
-# Structure your Dockerfile so that ${SANDBOX_REPO_DIR} can serve as the project root.
-ENTRYPOINT ["sleep", "infinity"]
-`;
-const CODEX_DOCKERFILE = `FROM node:22-bookworm
-
-# Install system dependencies
-RUN apt-get update && apt-get install -y \\
-  git \\
-  curl \\
-  jq \\
-  && rm -rf /var/lib/apt/lists/*
-
-{{ISSUE_TRACKER_TOOLS}}
-
-# Build-args for UID/GID alignment: sandcastle docker build-image
-# defaults these to the host user's UID/GID so image-built files
-# and bind-mounted files share an owner without runtime chown.
-ARG AGENT_UID=1000
-ARG AGENT_GID=1000
-
-# Rename the base image's "node" user to "agent" and align UID/GID.
-RUN groupmod -o -g $AGENT_GID node && usermod -o -u $AGENT_UID -g $AGENT_GID -d /home/agent -m -l agent node
-
-# Install Codex CLI (run as root before USER agent)
-RUN npm install -g @openai/codex
-
-USER \${AGENT_UID}:\${AGENT_GID}
-
-WORKDIR /home/agent
-
-# In worktree sandbox mode, Sandcastle bind-mounts the git worktree at ${SANDBOX_REPO_DIR}
-# and overrides the working directory to ${SANDBOX_REPO_DIR} at container start.
-# Structure your Dockerfile so that ${SANDBOX_REPO_DIR} can serve as the project root.
-ENTRYPOINT ["sleep", "infinity"]
-`;
-const CURSOR_DOCKERFILE = `FROM node:22-bookworm
-
-# Install system dependencies
-RUN apt-get update && apt-get install -y \\
-  git \\
-  curl \\
-  jq \\
-  && rm -rf /var/lib/apt/lists/*
-
-{{ISSUE_TRACKER_TOOLS}}
-
-# Build-args for UID/GID alignment: sandcastle docker build-image
-# defaults these to the host user's UID/GID so image-built files
-# and bind-mounted files share an owner without runtime chown.
-ARG AGENT_UID=1000
-ARG AGENT_GID=1000
-
-# Rename the base image's "node" user to "agent" and align UID/GID.
-RUN groupmod -g $AGENT_GID node && usermod -u $AGENT_UID -g $AGENT_GID -d /home/agent -m -l agent node
-USER \${AGENT_UID}:\${AGENT_GID}
-
-# Install Cursor Agent CLI
-RUN curl https://cursor.com/install -fsS | bash
-
-# Add Cursor CLI to PATH
-ENV PATH="/home/agent/.local/bin:$PATH"
-
-WORKDIR /home/agent
-
-# In worktree sandbox mode, Sandcastle bind-mounts the git worktree at ${SANDBOX_REPO_DIR}
-# and overrides the working directory to ${SANDBOX_REPO_DIR} at container start.
-# Structure your Dockerfile so that ${SANDBOX_REPO_DIR} can serve as the project root.
-ENTRYPOINT ["sleep", "infinity"]
-`;
-const OPENCODE_DOCKERFILE = `FROM node:22-bookworm
-
-# Install system dependencies
-RUN apt-get update && apt-get install -y \\
-  git \\
-  curl \\
-  jq \\
-  && rm -rf /var/lib/apt/lists/*
-
-{{ISSUE_TRACKER_TOOLS}}
-
-# Build-args for UID/GID alignment: sandcastle docker build-image
-# defaults these to the host user's UID/GID so image-built files
-# and bind-mounted files share an owner without runtime chown.
-ARG AGENT_UID=1000
-ARG AGENT_GID=1000
-
-# Rename the base image's "node" user to "agent" and align UID/GID.
-RUN groupmod -o -g $AGENT_GID node && usermod -o -u $AGENT_UID -g $AGENT_GID -d /home/agent -m -l agent node
-
-# Install OpenCode CLI (run as root before USER agent)
-RUN npm install -g opencode-ai@latest
-
-USER \${AGENT_UID}:\${AGENT_GID}
-
-WORKDIR /home/agent
-
-# In worktree sandbox mode, Sandcastle bind-mounts the git worktree at \${SANDBOX_REPO_DIR}
-# and overrides the working directory to \${SANDBOX_REPO_DIR} at container start.
-# Structure your Dockerfile so that \${SANDBOX_REPO_DIR} can serve as the project root.
-ENTRYPOINT ["sleep", "infinity"]
-`;
-const COPILOT_DOCKERFILE = `FROM node:22-bookworm
-
-# Install system dependencies
-RUN apt-get update && apt-get install -y \\
-  git \\
-  curl \\
-  jq \\
-  && rm -rf /var/lib/apt/lists/*
-
-{{ISSUE_TRACKER_TOOLS}}
-
-# Build-args for UID/GID alignment: sandcastle docker build-image
-# defaults these to the host user's UID/GID so image-built files
-# and bind-mounted files share an owner without runtime chown.
-ARG AGENT_UID=1000
-ARG AGENT_GID=1000
-
-# Rename the base image's "node" user to "agent" and align UID/GID.
-RUN groupmod -o -g $AGENT_GID node && usermod -o -u $AGENT_UID -g $AGENT_GID -d /home/agent -m -l agent node
-
-# Install GitHub Copilot CLI (run as root before USER agent)
-RUN npm install -g @github/copilot
-
-USER \${AGENT_UID}:\${AGENT_GID}
-
-WORKDIR /home/agent
-
-# In worktree sandbox mode, Sandcastle bind-mounts the git worktree at \${SANDBOX_REPO_DIR}
-# and overrides the working directory to \${SANDBOX_REPO_DIR} at container start.
-# Structure your Dockerfile so that \${SANDBOX_REPO_DIR} can serve as the project root.
-ENTRYPOINT ["sleep", "infinity"]
-`;
-const AGENT_REGISTRY = [
-    {
-        name: "claude-code",
-        label: "Claude Code",
-        defaultModel: "claude-opus-4-7",
-        factoryImport: "claudeCode",
-        dockerfileTemplate: CLAUDE_CODE_DOCKERFILE,
-        envExample: `# Anthropic API key
-# If you want to use your Claude subscription instead of an API key, see https://github.com/mattpocock/sandcastle/issues/191
-ANTHROPIC_API_KEY=`,
-        setupCommand: `claude "$(cat ${SETUP_ISSUE_TRACKER_PATH})"`,
-    },
-    {
-        name: "pi",
-        label: "Pi",
-        defaultModel: "claude-sonnet-4-6",
-        factoryImport: "pi",
-        dockerfileTemplate: PI_DOCKERFILE,
-        envExample: `# Anthropic API key
-ANTHROPIC_API_KEY=`,
-        setupCommand: `pi "$(cat ${SETUP_ISSUE_TRACKER_PATH})"`,
-    },
-    {
-        name: "codex",
-        label: "Codex",
-        defaultModel: "gpt-5.4-mini",
-        factoryImport: "codex",
-        dockerfileTemplate: CODEX_DOCKERFILE,
-        envExample: `# OpenAI API key
-OPENAI_KEY=`,
-        setupCommand: `codex "$(cat ${SETUP_ISSUE_TRACKER_PATH})"`,
-    },
-    {
-        name: "cursor",
-        label: "Cursor",
-        defaultModel: "composer-2",
-        factoryImport: "cursor",
-        dockerfileTemplate: CURSOR_DOCKERFILE,
-        envExample: `# Cursor API key (recommended)
-# You can also pass --api-key directly to the agent CLI.
-CURSOR_API_KEY=`,
-        setupCommand: `agent "$(cat ${SETUP_ISSUE_TRACKER_PATH})"`,
-    },
-    {
-        name: "opencode",
-        label: "OpenCode",
-        defaultModel: "opencode/big-pickle",
-        factoryImport: "opencode",
-        dockerfileTemplate: OPENCODE_DOCKERFILE,
-        envExample: `# OpenCode API key
-OPENCODE_API_KEY=`,
-        setupCommand: `opencode -p "$(cat ${SETUP_ISSUE_TRACKER_PATH})"`,
-    },
-    {
-        name: "copilot",
-        label: "GitHub Copilot CLI",
-        defaultModel: "claude-sonnet-4.5",
-        factoryImport: "copilot",
-        dockerfileTemplate: COPILOT_DOCKERFILE,
-        envExample: `# GitHub token with the "Copilot Requests" permission
-# (a fine-grained PAT, or any token from \`gh auth login\`).
-# COPILOT_GITHUB_TOKEN takes precedence over GH_TOKEN and GITHUB_TOKEN.
-GITHUB_TOKEN=`,
-        setupCommand: `copilot -i "$(cat ${SETUP_ISSUE_TRACKER_PATH})"`,
-    },
-];
-export const listAgents = () => AGENT_REGISTRY;
-const GITHUB_CLI_TOOLS = `# Install GitHub CLI
-RUN curl -fsSL https://cli.github.com/packages/githubcli-archive-keyring.gpg \\
-  | dd of=/usr/share/keyrings/githubcli-archive-keyring.gpg \\
-  && echo "deb [arch=$(dpkg --print-architecture) signed-by=/usr/share/keyrings/githubcli-archive-keyring.gpg] https://cli.github.com/packages stable main" \\
-  | tee /etc/apt/sources.list.d/github-cli.list > /dev/null \\
-  && apt-get update && apt-get install -y gh \\
-  && rm -rf /var/lib/apt/lists/*`;
-const BEADS_TOOLS = `# Install system dependencies for Beads
-RUN apt-get update && apt-get install -y \\
-  dpkg-dev \\
-  libicu72 \\
-  && rm -rf /var/lib/apt/lists/* \\
-  && ARCH_DIR=$(dpkg-architecture -qDEB_HOST_MULTIARCH) \\
-  && for lib in /usr/lib/$ARCH_DIR/libicu*.so.72; do \\
-       ln -s "$lib" "\${lib%.72}.74"; \\
-     done
-
-RUN curl -fsSL https://raw.githubusercontent.com/steveyegge/beads/main/scripts/install.sh | bash
-
-RUN corepack enable`;
-// Sentinels baked into the scaffold for the `custom` issue tracker. The
-// project ships deliberately broken-until-configured; the setup agent finds
-// and replaces these markers in place (see SETUP_ISSUE_TRACKER.md). Defined as
-// shared constants so the registry entry and the setup doc stay in sync.
-const CUSTOM_LIST_TASKS_SENTINEL = `echo 'No issue tracker configured — run ${SETUP_ISSUE_TRACKER_PATH} through your coding agent.' >&2; exit 1`;
-const CUSTOM_VIEW_TASK_MARKER = `<view command — see ${SETUP_ISSUE_TRACKER_PATH}>`;
-const CUSTOM_CLOSE_TASK_MARKER = `<close command — see ${SETUP_ISSUE_TRACKER_PATH}>`;
-const CUSTOM_TRACKER_TOOLS = `# TODO: install your issue tracker's CLI here. See ${SETUP_ISSUE_TRACKER_PATH}`;
-const CUSTOM_ENV_EXAMPLE = `# TODO: add any env vars your issue tracker needs (e.g. an API token).
-# See ${SETUP_ISSUE_TRACKER_PATH}`;
-const ISSUE_TRACKER_REGISTRY = [
-    {
-        name: "github-issues",
-        label: "GitHub Issues",
-        templateArgs: {
-            LIST_TASKS_COMMAND: `gh issue list --state open --label Sandcastle --limit 100 --json number,title,body,labels,comments --jq '[.[] | {number, title, body, labels: [.labels[].name], comments: [.comments[].body]}]'`,
-            VIEW_TASK_COMMAND: "gh issue view <ID>",
-            CLOSE_TASK_COMMAND: `gh issue close <ID> --comment "Completed by Sandcastle"`,
-            ISSUE_TRACKER_TOOLS: GITHUB_CLI_TOOLS,
-        },
-        envExample: `# GitHub personal access token — the agent uses it to read and manage GitHub Issues
-# Create a fine-grained token: https://github.com/settings/personal-access-tokens/new
-# Required repository permissions: Issues (Read and write) and Metadata (Read)
-GH_TOKEN=`,
-    },
-    {
-        name: "beads",
-        label: "Beads",
-        templateArgs: {
-            LIST_TASKS_COMMAND: "bd ready --json",
-            VIEW_TASK_COMMAND: "bd show <ID>",
-            CLOSE_TASK_COMMAND: `bd close <ID> --reason="Completed by Sandcastle"`,
-            ISSUE_TRACKER_TOOLS: BEADS_TOOLS,
-        },
-        envExample: "",
-    },
-    {
-        name: "custom",
-        label: "Custom",
-        templateArgs: {
-            // The only real shell expression: PromptPreprocessor fails the run on a
-            // non-zero exit and surfaces stderr, so this is the single enforcement
-            // point that keeps the scaffold broken until the user configures it.
-            LIST_TASKS_COMMAND: CUSTOM_LIST_TASKS_SENTINEL,
-            // Inline text markers — replaced by the setup agent, never executed.
-            VIEW_TASK_COMMAND: CUSTOM_VIEW_TASK_MARKER,
-            CLOSE_TASK_COMMAND: CUSTOM_CLOSE_TASK_MARKER,
-            ISSUE_TRACKER_TOOLS: CUSTOM_TRACKER_TOOLS,
-        },
-        envExample: CUSTOM_ENV_EXAMPLE,
-    },
-];
-export const listIssueTrackers = () => ISSUE_TRACKER_REGISTRY;
-export const getIssueTracker = (name) => ISSUE_TRACKER_REGISTRY.find((b) => b.name === name);
-export const getAgent = (name) => AGENT_REGISTRY.find((a) => a.name === name);
-const SANDBOX_PROVIDER_REGISTRY = [
-    {
-        name: "docker",
-        label: "Docker",
-        containerfileName: "Dockerfile",
-        cliNamespace: "docker",
-    },
-    {
-        name: "podman",
-        label: "Podman",
-        containerfileName: "Containerfile",
-        cliNamespace: "podman",
-    },
-];
-export const listSandboxProviders = () => SANDBOX_PROVIDER_REGISTRY;
-export const getSandboxProvider = (name) => SANDBOX_PROVIDER_REGISTRY.find((p) => p.name === name);
-// ---------------------------------------------------------------------------
-// Next steps
-// ---------------------------------------------------------------------------
-export function getNextStepsLines(template, mainFilename, issueTracker, agent) {
-    // The custom issue tracker scaffolds a broken-until-configured project, so
-    // its next steps are about running the setup prompt — not the template's
-    // normal "set env vars and go" flow. This branch wins over template-specific
-    // steps regardless of the chosen template.
-    if (issueTracker.name === "custom") {
-        return [
-            "Next steps:",
-            "1. Your custom issue tracker isn't wired up yet — runs hard-fail until you configure it.",
-            `2. Feed the setup prompt to ${agent.label} on your host to finish wiring it up:`,
-            `   ${agent.setupCommand}`,
-            `   (Runs on the host — you need the ${agent.label} CLI installed locally, since the sandbox image isn't built yet.)`,
-            `3. Follow .sandcastle/${SETUP_ISSUE_TRACKER_DOC} to edit the scaffolded files in place, build the image, and verify.`,
-        ];
-    }
-    if (template === "blank") {
-        return [
-            "Next steps:",
-            `1. Set the required env vars in .sandcastle/.env (see .sandcastle/.env.example)`,
-            "   If you want to use your Claude subscription instead of an API key, see https://github.com/mattpocock/sandcastle/issues/191",
-            "2. Read and customize .sandcastle/prompt.md to describe what you want the agent to do",
-            `3. Customize .sandcastle/${mainFilename} — it uses the JS API (\`run()\`) to control how the agent runs`,
-            `4. Add "sandcastle": "npx tsx .sandcastle/${mainFilename}" to your package.json scripts`,
-            "5. Run `npm run sandcastle` to start the agent",
-        ];
-    }
-    else {
-        const hasReviewer = template.includes("review");
-        const usesPlanSchema = template.includes("planner");
-        let step = 1;
-        const lines = [
-            "Next steps:",
-            `${step++}. Set the required env vars in .sandcastle/.env (see .sandcastle/.env.example)`,
-            "   If you want to use your Claude subscription instead of an API key, see https://github.com/mattpocock/sandcastle/issues/191",
-            `${step++}. Add "sandcastle": "npx tsx .sandcastle/${mainFilename}" to your package.json scripts`,
-            `${step++}. Templates use \`copyToWorktree: ["node_modules"]\` to copy your host node_modules into the sandbox for fast startup — the \`npm install\` in the onSandboxReady hook is a safety net for platform-specific binaries. Adjust both if you use a different package manager`,
-        ];
-        if (usesPlanSchema) {
-            lines.push(`${step++}. Install a schema validator for the planner's \`<plan>\` output — the template uses Zod (\`npm install zod\`), but Valibot, ArkType, or any Standard Schema library works (https://standardschema.dev)`);
-        }
-        lines.push(`${step++}. Read and customize the prompt files in .sandcastle/ — they shape what the agent does`);
-        if (hasReviewer) {
-            lines.push(`${step++}. Customize .sandcastle/CODING_STANDARDS.md with your project's standards — the reviewer agent loads it during review`);
-        }
-        lines.push(`${step++}. Run \`npm run sandcastle\` to start the agent`);
-        return lines;
-    }
-}
-// ---------------------------------------------------------------------------
-// Scaffolding helpers
-// ---------------------------------------------------------------------------
-function getTemplatesDir() {
-    const thisFile = fileURLToPath(import.meta.url);
-    return join(dirname(thisFile), "templates");
-}
-const getTemplateDir = (templateName) => Effect.gen(function* () {
-    const template = TEMPLATES.find((t) => t.name === templateName);
-    if (!template) {
-        const names = TEMPLATES.map((t) => t.name).join(", ");
-        yield* Effect.fail(new Error(`Unknown template: "${templateName}". Available: ${names}`));
-    }
-    return join(getTemplatesDir(), templateName);
-});
-const COMPILED_FILE_EXTENSIONS = [
-    ".js",
-    ".js.map",
-    ".d.ts",
-    ".d.ts.map",
-    ".mjs",
-    ".mjs.map",
-    ".d.mts",
-    ".d.mts.map",
-];
-const copyTemplateFiles = (templateDir, destDir, mainFilename) => Effect.gen(function* () {
-    const fs = yield* FileSystem.FileSystem;
-    const files = yield* fs
-        .readDirectory(templateDir)
-        .pipe(Effect.mapError((e) => new Error(e.message)));
-    yield* Effect.all(files
-        .filter((f) => f !== "template.json" &&
-        f !== ".env.example" &&
-        !COMPILED_FILE_EXTENSIONS.some((ext) => f.endsWith(ext)))
-        .map((f) => {
-        const destName = f === "main.mts" ? mainFilename : f;
-        return fs
-            .copyFile(join(templateDir, f), join(destDir, destName))
-            .pipe(Effect.mapError((e) => new Error(e.message)));
-    }), { concurrency: "unbounded" });
-});
-/**
- * Replace the agent factory and sandbox provider in a scaffolded main.ts.
- *
- * Templates use `claudeCode` as the default agent factory and `docker` as the
- * default sandbox provider. When a different agent, model, or sandbox provider
- * is selected, this function rewrites the imports and factory calls.
- */
-const rewriteMainTs = (configDir, agent, model, sandboxProvider, mainFilename) => Effect.gen(function* () {
-    const fs = yield* FileSystem.FileSystem;
-    const mainTsPath = join(configDir, mainFilename);
-    const exists = yield* fs
-        .exists(mainTsPath)
-        .pipe(Effect.mapError((e) => new Error(e.message)));
-    if (!exists)
-        return;
-    let content = yield* fs
-        .readFileString(mainTsPath)
-        .pipe(Effect.mapError((e) => new Error(e.message)));
-    // Templates use main.mts as the canonical filename in comments.
-    // When the target is main.ts, rewrite those references.
-    if (mainFilename === "main.ts") {
-        content = content.replace(/main\.mts/g, "main.ts");
-    }
-    // Replace factory function name in imports (e.g. claudeCode → pi)
-    // and all factory calls with the correct model.
-    // Templates always use claudeCode as the placeholder factory.
-    content = content.replace(/\bclaudeCode\b/g, agent.factoryImport);
-    // Replace model strings in factory calls: factoryImport("any-model")
-    const factoryCallRe = new RegExp(`${agent.factoryImport}\\(["']([^"']+)["']\\)`, "g");
-    content = content.replace(factoryCallRe, `${agent.factoryImport}("${model}")`);
-    // Replace the sandbox provider. Templates always use `docker` as the
-    // placeholder, where the registry name doubles as both the factory function
-    // name and the `/sandboxes/<name>` import subpath segment. A single
-    // case-sensitive word-boundary replace therefore rewrites the named import,
-    // the import subpath, and every factory call site — and is a no-op when
-    // docker is selected.
-    content = content.replace(/\bdocker\b/g, sandboxProvider.name);
-    yield* fs
-        .writeFileString(mainTsPath, content)
-        .pipe(Effect.mapError((e) => new Error(e.message)));
-});
-/**
- * When the user opted out of the Sandcastle label, strip ` --label Sandcastle`
- * from all `.md` files in the scaffolded config directory so that `gh issue list`
- * commands work without a label filter.
- */
-const rewritePromptFiles = (configDir) => Effect.gen(function* () {
-    const fs = yield* FileSystem.FileSystem;
-    const files = yield* fs
-        .readDirectory(configDir)
-        .pipe(Effect.mapError((e) => new Error(e.message)));
-    const mdFiles = files.filter((f) => f.endsWith(".md"));
-    yield* Effect.all(mdFiles.map((f) => Effect.gen(function* () {
-        const filePath = join(configDir, f);
-        const content = yield* fs
-            .readFileString(filePath)
-            .pipe(Effect.mapError((e) => new Error(e.message)));
-        const updated = content.replace(/ --label Sandcastle/g, "");
-        if (updated !== content) {
-            yield* fs
-                .writeFileString(filePath, updated)
-                .pipe(Effect.mapError((e) => new Error(e.message)));
-        }
-    })), { concurrency: "unbounded" });
-});
-/** Text file extensions eligible for `{{KEY}}` template argument substitution. */
-const TEXT_FILE_EXTENSIONS = new Set([
-    ".md",
-    ".txt",
-    ".env",
-    ".example",
-    // Dockerfile / Containerfile have no extension — handled by name check below
-]);
-const isTextFile = (filename) => {
-    if (filename === "Dockerfile" ||
-        filename === "Containerfile" ||
-        filename === ".gitignore")
-        return true;
-    const dotIdx = filename.lastIndexOf(".");
-    if (dotIdx === -1)
-        return false;
-    return TEXT_FILE_EXTENSIONS.has(filename.slice(dotIdx));
-};
-/**
- * Replace `{{KEY}}` template arguments from the issue tracker's
- * `templateArgs` map in all text files in the scaffolded config directory.
- */
-const substituteTemplateArgs = (configDir, issueTracker) => Effect.gen(function* () {
-    const fs = yield* FileSystem.FileSystem;
-    const files = yield* fs
-        .readDirectory(configDir)
-        .pipe(Effect.mapError((e) => new Error(e.message)));
-    const textFiles = files.filter(isTextFile);
-    yield* Effect.all(textFiles.map((f) => Effect.gen(function* () {
-        const filePath = join(configDir, f);
-        let content = yield* fs
-            .readFileString(filePath)
-            .pipe(Effect.mapError((e) => new Error(e.message)));
-        const original = content;
-        for (const [key, value] of Object.entries(issueTracker.templateArgs)) {
-            content = content.replace(new RegExp(`\\{\\{${key}\\}\\}`, "g"), value);
-        }
-        if (content !== original) {
-            yield* fs
-                .writeFileString(filePath, content)
-                .pipe(Effect.mapError((e) => new Error(e.message)));
-        }
-    })), { concurrency: "unbounded" });
-});
-/**
- * Build the `SETUP_ISSUE_TRACKER.md` prompt scaffolded for the `custom` issue
- * tracker. It addresses the user's coding agent and walks it through wiring up
- * the tracker by editing the scaffolded files in place. The build command is
- * provider-parameterized so it names the actual CLI namespace (docker/podman).
- */
-const buildSetupIssueTrackerDoc = (cliNamespace) => `# Set up your custom issue tracker
-
-You are a coding agent. Finish wiring up the **custom issue tracker** for this Sandcastle project. It was scaffolded in a deliberately broken-until-configured state: until you complete the steps below, every Sandcastle run hard-fails with a pointer back to this file.
-
-## Goal
-
-Wire up the issue tracker so the scaffolded prompts can **list**, **view**, and **close** tasks. There is no runtime abstraction to implement — the tracker commands are baked into the scaffolded files, so you edit those files **in place**.
-
-## 1. Interview the user
-
-Ask the user:
-
-- Which issue tracker do they use (e.g. Jira, Linear, a GitHub repo other than this one, an internal API)?
-- How should the sandbox authenticate — a CLI that is already logged in, or an API token? If a token, what is the environment variable name?
-
-## 2. Produce three commands
-
-Work out, together with the user, the shell commands for:
-
-- **list** — print all open tasks **as JSON** (match the shape the built-in trackers emit: an array of objects, each with at least an id/number, title, and body). This is what the agent reads at the start of every iteration.
-- **view** \`<ID>\` — show a single task by id.
-- **close** \`<ID>\` — close a single task by id.
-
-## 3. Edit the scaffolded files in place
-
-- **Dockerfile / Containerfile** — replace the line
-
-  \`\`\`
-  ${CUSTOM_TRACKER_TOOLS}
-  \`\`\`
-
-  with the install steps for your tracker's CLI (if it needs one).
-
-- **Prompt files (\`.sandcastle/*.md\`)** — replace the sentinel
-
-  \`\`\`
-  ${CUSTOM_LIST_TASKS_SENTINEL}
-  \`\`\`
-
-  with your **list** command. In the prompt file the sentinel sits inside a Sandcastle **shell expression** — a leading \`!\` followed by the command in backticks — whose output is injected into the prompt before each run. Keep that \`!\` and the surrounding backticks; replace only the command between them, and **remove the \`exit 1\`** (leaving it keeps every run hard-failing). Then replace the \`${CUSTOM_VIEW_TASK_MARKER}\` and \`${CUSTOM_CLOSE_TASK_MARKER}\` markers with your **view** and **close** commands.
-
-- **\`.env.example\`** — replace the \`# TODO\` block with the real env var(s) your tracker needs, then tell the user to set them in \`.sandcastle/.env\`.
-
-## 4. Build the image
-
-Once the files are wired up, build the sandbox image:
-
-\`\`\`
-sandcastle ${cliNamespace} build-image
-\`\`\`
-
-## 5. Verify
-
-Run your **list** command inside the built image and confirm it returns the open tasks as JSON. If it errors, fix the command or the auth and rebuild.
-`;
-/**
- * Detect whether the project's package.json has `"type": "module"`.
- * If so, we can use plain `.ts`; otherwise we use `.mts` to ensure ESM.
- */
-const detectMainFilename = (repoDir) => Effect.gen(function* () {
-    const fs = yield* FileSystem.FileSystem;
-    const pkgPath = join(repoDir, "package.json");
-    const exists = yield* fs
-        .exists(pkgPath)
-        .pipe(Effect.orElseSucceed(() => false));
-    if (!exists)
-        return "main.mts";
-    const content = yield* fs
-        .readFileString(pkgPath)
-        .pipe(Effect.orElseSucceed(() => ""));
-    try {
-        const pkg = JSON.parse(content);
-        return pkg["type"] === "module" ? "main.ts" : "main.mts";
-    }
-    catch {
-        return "main.mts";
-    }
-});
-export const scaffold = (repoDir, options) => Effect.gen(function* () {
-    const { agent, model, templateName = "blank", createLabel = true, issueTracker = ISSUE_TRACKER_REGISTRY[0], // default: github-issues
-    sandboxProvider = SANDBOX_PROVIDER_REGISTRY[0], // default: docker
-     } = options;
-    const fs = yield* FileSystem.FileSystem;
-    const configDir = join(repoDir, ".sandcastle");
-    const exists = yield* fs
-        .exists(configDir)
-        .pipe(Effect.mapError((e) => new Error(e.message)));
-    if (exists) {
-        yield* Effect.fail(new Error(".sandcastle/ directory already exists. Remove it first if you want to re-initialize."));
-    }
-    const mainFilename = yield* detectMainFilename(repoDir);
-    yield* fs
-        .makeDirectory(configDir, { recursive: false })
-        .pipe(Effect.mapError((e) => new Error(e.message)));
-    const templateDir = yield* getTemplateDir(templateName);
-    // Build .env.example from agent + issue tracker env blocks
-    const envExampleParts = [agent.envExample];
-    if (issueTracker.envExample) {
-        envExampleParts.push(issueTracker.envExample);
-    }
-    const envExampleContent = envExampleParts.join("\n") + "\n";
-    yield* Effect.all([
-        fs
-            .writeFileString(join(configDir, sandboxProvider.containerfileName), agent.dockerfileTemplate)
-            .pipe(Effect.mapError((e) => new Error(e.message))),
-        fs
-            .writeFileString(join(configDir, ".gitignore"), GITIGNORE)
-            .pipe(Effect.mapError((e) => new Error(e.message))),
-        fs
-            .writeFileString(join(configDir, ".env.example"), envExampleContent)
-            .pipe(Effect.mapError((e) => new Error(e.message))),
-        copyTemplateFiles(templateDir, configDir, mainFilename),
-    ], { concurrency: "unbounded" });
-    // Rewrite main file with the selected agent factory, model, and sandbox provider
-    yield* rewriteMainTs(configDir, agent, model, sandboxProvider, mainFilename);
-    // Replace issue tracker template arguments in all text files (must run before label stripping)
-    yield* substituteTemplateArgs(configDir, issueTracker);
-    // Strip --label Sandcastle from prompt files when the user declined label creation
-    if (!createLabel) {
-        yield* rewritePromptFiles(configDir);
-    }
-    // For the custom issue tracker, drop the setup prompt the user feeds to
-    // their coding agent. Written after substituteTemplateArgs so it isn't
-    // clobbered and references the resolved sentinel markers the agent finds
-    // (not the {{KEY}} names, which are gone by now).
-    if (issueTracker.name === "custom") {
-        yield* fs
-            .writeFileString(join(configDir, SETUP_ISSUE_TRACKER_DOC), buildSetupIssueTrackerDoc(sandboxProvider.cliNamespace))
-            .pipe(Effect.mapError((e) => new Error(e.message)));
-    }
-    return { mainFilename };
-});
-//# sourceMappingURL=InitService.js.map