@shahmarasy/prodo 0.1.3 → 0.1.5

package/src/{cli.ts → cli/index.ts}

@@ -1,319 +1,459 @@
-import { Command } from "commander";
-import { createHash } from "node:crypto";
-import fs from "node:fs/promises";
-import path from "node:path";
-import { loadAgentCommandSet, resolveAgent } from "./agents";
-import { resolveAi, type SupportedAi } from "./agent-command-installer";
-import { listArtifactTypes } from "./artifact-registry";
-import { generateArtifact } from "./artifacts";
-import { runDoctor } from "./doctor";
-import { UserError } from "./errors";
-import { runHookPhase } from "./hook-executor";
-import { runInit } from "./init";
-import { finishInitInteractive, gatherInitSelections } from "./init-tui";
-import { runNormalize } from "./normalize";
-import { briefPath } from "./paths";
-import { type ArtifactType } from "./types";
-import { fileExists } from "./utils";
-import { runValidate } from "./validate";
-import { readCliVersion } from "./version";
-
-type RunOptions = {
-  forcedCommand?: string;
-  cwd?: string;
-  argv?: string[];
-  log?: (message: string) => void;
-  error?: (message: string) => void;
-};
-
-const dynamicImport = new Function("specifier", "return import(specifier)") as (
-  specifier: string
-) => Promise<unknown>;
-
-function mapForcedCommand(forcedCommand: string): ArtifactType | "init" | "validate" | "normalize" | undefined {
-  if (forcedCommand === "prodo-init") return "init";
-  if (forcedCommand === "prodo-validate") return "validate";
-  if (forcedCommand === "prodo-normalize") return "normalize";
-  if (forcedCommand === "prodo-prd") return "prd";
-  if (forcedCommand === "prodo-workflow") return "workflow";
-  if (forcedCommand === "prodo-wireframe") return "wireframe";
-  if (forcedCommand === "prodo-stories") return "stories";
-  if (forcedCommand === "prodo-techspec") return "techspec";
-  return undefined;
-}
-
-async function runArtifactCommand(
-  type: ArtifactType,
-  opts: { from?: string; out?: string; agent?: string },
-  cwd: string,
-  log: (message: string) => void,
-  options?: { suggestValidate?: boolean }
-): Promise<void> {
-  await runHookPhase(cwd, `before_${type}`, log);
-  const agent = resolveAgent(opts.agent);
-  const file = await generateArtifact({
-    artifactType: type,
-    cwd,
-    normalizedBriefOverride: opts.from,
-    outPath: opts.out,
-    agent
-  });
-  const agentMsg = agent ? ` [agent=${agent}]` : "";
-  log(`${type.toUpperCase()} generated${agentMsg}: ${file}`);
-  if (options?.suggestValidate !== false) {
-    log("Tip: run `prodo validate` to check cross-artifact consistency.");
-  }
-  await runHookPhase(cwd, `after_${type}`, log);
-}
-
-type BriefSnapshot = {
-  hash: string;
-  mtimeMs: number;
-  size: number;
-};
-
-async function snapshotBrief(cwd: string): Promise<BriefSnapshot | null> {
-  const file = briefPath(cwd);
-  if (!(await fileExists(file))) return null;
-  const [raw, stat] = await Promise.all([fs.readFile(file), fs.stat(file)]);
-  return {
-    hash: createHash("sha256").update(raw).digest("hex"),
-    mtimeMs: stat.mtimeMs,
-    size: stat.size
-  };
-}
-
-async function withBriefReadOnlyGuard(cwd: string, task: () => Promise<void>): Promise<void> {
-  const before = await snapshotBrief(cwd);
-  await task();
-  const after = await snapshotBrief(cwd);
-  if (!before) return;
-  if (!after) {
-    throw new UserError("Input file `brief.md` was removed during execution. Input files are read-only.");
-  }
-  if (before.hash !== after.hash || before.size !== after.size || before.mtimeMs !== after.mtimeMs) {
-    throw new UserError("Input file `brief.md` was modified during execution. Input files are read-only.");
-  }
-}
-
-export async function runCli(options: RunOptions = {}): Promise<number> {
-  const cwd = options.cwd ?? process.cwd();
-  const argv = options.argv ?? process.argv;
-  const out = options.log ?? console.log;
-  const err = options.error ?? console.error;
-  const forced = options.forcedCommand ? mapForcedCommand(options.forcedCommand) : undefined;
-
-  const program = new Command();
-  const version = await readCliVersion(cwd);
-  program
-    .name("prodo")
-    .description("CLI-first, prompt-powered product artifact kit")
-    .version(`prodo ${version}`, "-v, --version", "Show Prodo version")
-    .showHelpAfterError();
-  const artifactTypes = await listArtifactTypes(cwd);
-
-  program
-    .command("init [target]")
-    .option("--ai <name>", "agent integration: codex | gemini-cli | claude-cli")
-    .option("--lang <code>", "document language (e.g. en, tr)")
-    .option("--preset <name>", "preset to install during initialization")
-    .action(async (target, opts) => {
-      const projectRoot = path.resolve(cwd, target ?? ".");
-      const selected = await gatherInitSelections({
-        projectRoot,
-        aiInput: opts.ai,
-        langInput: opts.lang
-      });
-      const selectedAi = selected.ai as SupportedAi | undefined;
-
-      if (selected.interactive) {
-        const clack = (await dynamicImport("@clack/prompts")) as typeof import("@clack/prompts");
-        const s = clack.spinner();
-        s.start("Scaffolding Prodo workspace...");
-        const result = await runInit(projectRoot, {
-          ai: selectedAi,
-          lang: selected.lang,
-          preset: opts.preset,
-          script: selected.script
-        });
-        s.stop("Scaffold complete.");
-        await finishInitInteractive({
-          projectRoot,
-          settingsPath: result.settingsPath,
-          ai: selectedAi,
-          lang: selected.lang
-        });
-        return;
-      }
-
-      const result = await runInit(projectRoot, {
-        ai: selectedAi,
-        lang: selected.lang,
-        preset: opts.preset,
-        script: selected.script
-      });
-      out(`Initialized Prodo scaffold at ${path.join(projectRoot, ".prodo")}`);
-      if (selectedAi) {
-        out(`Agent command set installed for ${selectedAi}.`);
-        out(`Installed ${result.installedAgentFiles.length} command files.`);
-        out("Agent workflow: edit brief.md, then run slash commands in your agent.");
-      } else {
-        out("No agent selected. Use `prodo generate` for end-to-end generation.");
-      }
-      out(`Settings file: ${result.settingsPath}`);
-      out("Next: edit brief.md.");
-    });
-
-  program
-    .command("generate")
-    .description("Run end-to-end pipeline: normalize -> generate artifacts -> validate")
-    .option("--agent <name>", "agent profile: codex | gemini-cli | claude-cli")
-    .option("--strict", "treat validation warnings as errors")
-    .option("--report <path>", "validation report output path")
-    .action(async (opts) => {
-      if (opts.agent) resolveAgent(opts.agent);
-      await withBriefReadOnlyGuard(cwd, async () => {
-        await runHookPhase(cwd, "before_normalize", out);
-        const normalizedPath = await runNormalize({ cwd });
-        out(`Normalized brief written to: ${normalizedPath}`);
-        await runHookPhase(cwd, "after_normalize", out);
-
-        for (const type of artifactTypes) {
-          await runArtifactCommand(type, { from: normalizedPath, agent: opts.agent }, cwd, out, {
-            suggestValidate: false
-          });
-        }
-
-        await runHookPhase(cwd, "before_validate", out);
-        const result = await runValidate(cwd, {
-          strict: Boolean(opts.strict),
-          report: opts.report
-        });
-        out(`Validation report written to: ${result.reportPath}`);
-        if (!result.pass) {
-          throw new UserError("Validation failed. Review report and fix issues.");
-        }
-        out("Generation pipeline completed. Validation passed.");
-        await runHookPhase(cwd, "after_validate", out);
-      });
-    });
-
-  program
-    .command("normalize", { hidden: true })
-    .description("Advanced: normalize brief without full pipeline")
-    .option("--brief <path>", "path to start brief markdown")
-    .option("--out <path>", "output normalized brief json path")
-    .option("--agent <name>", "agent profile: codex | gemini-cli | claude-cli")
-    .action(async (opts: { brief?: string; out?: string; agent?: string }) => {
-      if (opts.agent) resolveAgent(opts.agent);
-      await withBriefReadOnlyGuard(cwd, async () => {
-        await runHookPhase(cwd, "before_normalize", out);
-        const outPath = await runNormalize({
-          cwd,
-          brief: opts.brief,
-          out: opts.out
-        });
-        out(`Normalized brief written to: ${outPath}`);
-        await runHookPhase(cwd, "after_normalize", out);
-      });
-    });
-
-  program
-    .command("doctor")
-    .alias("check")
-    .description("Check local environment and toolchain readiness")
-    .action(async () => {
-      await runDoctor(cwd, out);
-    });
-
-  for (const type of artifactTypes) {
-    program
-      .command(type, { hidden: true })
-      .description(`Advanced: generate only ${type} artifact`)
-      .option("--from <path>", "path to normalized-brief.json")
-      .option("--out <path>", "output file path")
-      .option("--agent <name>", "agent profile: codex | gemini-cli | claude-cli")
-      .action(async (opts: { from?: string; out?: string; agent?: string }) => {
-        await withBriefReadOnlyGuard(cwd, async () => {
-          await runArtifactCommand(type, opts, cwd, out);
-        });
-      });
-  }
-
-  program
-    .command("agent-commands", { hidden: true })
-    .requiredOption("--agent <name>", "agent profile: codex | gemini-cli | claude-cli")
-    .action(async (opts: { agent: string }) => {
-      const agent = resolveAgent(opts.agent);
-      if (!agent) throw new UserError("Agent is required.");
-      const set = await loadAgentCommandSet(cwd, agent);
-      out(`Agent: ${set.agent}`);
-      if (set.description) out(`Description: ${set.description}`);
-      out("");
-      out("Recommended sequence:");
-      for (const item of set.recommended_sequence ?? []) {
-        out(`- ${item.command}: ${item.purpose}`);
-      }
-      if (set.artifact_shortcuts) {
-        out("");
-        out("Artifact shortcuts:");
-        for (const [key, command] of Object.entries(set.artifact_shortcuts)) {
-          out(`- ${key}: ${command}`);
-        }
-      }
-    });
-
-  program
-    .command("validate", { hidden: true })
-    .description("Advanced: run validation only")
-    .option("--strict", "treat warnings as errors")
-    .option("--report <path>", "report output path")
-    .option("--agent <name>", "agent profile: codex | gemini-cli | claude-cli")
-    .action(async (opts: { strict?: boolean; report?: string; agent?: string }) => {
-      if (opts.agent) resolveAgent(opts.agent);
-      await withBriefReadOnlyGuard(cwd, async () => {
-        await runHookPhase(cwd, "before_validate", out);
-        const result = await runValidate(cwd, {
-          strict: Boolean(opts.strict),
-          report: opts.report
-        });
-        out(`Validation report written to: ${result.reportPath}`);
-        if (!result.pass) {
-          throw new UserError("Validation failed. Review report and fix issues.");
-        }
-        out("Validation passed.");
-        await runHookPhase(cwd, "after_validate", out);
-      });
-    });
-
-  try {
-    if (forced) {
-      if (forced === "init") {
-        await program.parseAsync(["node", "prodo", "init", ...argv.slice(2)]);
-      } else if (forced === "normalize") {
-        await program.parseAsync(["node", "prodo", "normalize", ...argv.slice(2)]);
-      } else if (forced === "validate") {
-        await program.parseAsync(["node", "prodo", "validate", ...argv.slice(2)]);
-      } else {
-        await program.parseAsync(["node", "prodo", forced, ...argv.slice(2)]);
-      }
-    } else {
-      await program.parseAsync(argv);
-    }
-    return 0;
-  } catch (error) {
-    if (error instanceof UserError) {
-      err(error.message);
-      return 1;
-    }
-    const unknown = error as Error;
-    err(unknown.message);
-    return 1;
-  }
-}
-
-if (require.main === module) {
-  runCli().then((code) => {
-    process.exitCode = code;
-  });
-}
+import { Command } from "commander";
+import { createHash } from "node:crypto";
+import fs from "node:fs/promises";
+import path from "node:path";
+import { loadAgentCommandSet, resolveAgent } from "./agent-ids";
+import { resolveAi, type SupportedAi } from "./agent-command-installer";
+import { listArtifactTypes } from "../core/artifact-registry";
+import { generateArtifact } from "../core/artifacts";
+import { runDoctor } from "./doctor";
+import { UserError } from "../core/errors";
+import { runHookPhase } from "../core/hook-executor";
+import { runInit } from "./init";
+import { finishInitInteractive, gatherInitSelections } from "./init-tui";
+import { runClean } from "../core/clean";
+import { buildFixProposal, applyFix, runFix } from "../core/fix";
+import { runNormalize } from "../core/normalize";
+import { runInteractiveNormalize } from "./normalize-interactive";
+import { confirmFixExecution, displayFixProposal, displayFixResult } from "./fix-tui";
+import { briefPath } from "../core/paths";
+import { type ArtifactType } from "../core/types";
+import { fileExists } from "../core/utils";
+import { runValidate } from "../core/validate";
+import { readCliVersion } from "../core/version";
+
+type RunOptions = {
+  forcedCommand?: string;
+  cwd?: string;
+  argv?: string[];
+  log?: (message: string) => void;
+  error?: (message: string) => void;
+};
+
+const dynamicImport = new Function("specifier", "return import(specifier)") as (
+  specifier: string
+) => Promise<unknown>;
+
+function mapForcedCommand(forcedCommand: string): ArtifactType | "init" | "validate" | "normalize" | "fix" | undefined {
+  if (forcedCommand === "prodo-init") return "init";
+  if (forcedCommand === "prodo-validate") return "validate";
+  if (forcedCommand === "prodo-normalize") return "normalize";
+  if (forcedCommand === "prodo-fix") return "fix";
+  if (forcedCommand === "prodo-prd") return "prd";
+  if (forcedCommand === "prodo-workflow") return "workflow";
+  if (forcedCommand === "prodo-wireframe") return "wireframe";
+  if (forcedCommand === "prodo-stories") return "stories";
+  if (forcedCommand === "prodo-techspec") return "techspec";
+  return undefined;
+}
+
+async function runArtifactCommand(
+  type: ArtifactType,
+  opts: { from?: string; out?: string; agent?: string; revisionType?: "default" | "fix" },
+  cwd: string,
+  log: (message: string) => void,
+  options?: { suggestValidate?: boolean }
+): Promise<void> {
+  await runHookPhase(cwd, `before_${type}`, log);
+  const agent = resolveAgent(opts.agent);
+  const file = await generateArtifact({
+    artifactType: type,
+    cwd,
+    normalizedBriefOverride: opts.from,
+    outPath: opts.out,
+    agent,
+    revisionType: opts.revisionType
+  });
+  const agentMsg = agent ? ` [agent=${agent}]` : "";
+  log(`${type.toUpperCase()} generated${agentMsg}: ${file}`);
+  if (options?.suggestValidate !== false) {
+    log("Tip: run `prodo validate` to check cross-artifact consistency.");
+  }
+  await runHookPhase(cwd, `after_${type}`, log);
+}
+
+type BriefSnapshot = {
+  hash: string;
+  mtimeMs: number;
+  size: number;
+};
+
+async function snapshotBrief(cwd: string): Promise<BriefSnapshot | null> {
+  const file = briefPath(cwd);
+  if (!(await fileExists(file))) return null;
+  const [raw, stat] = await Promise.all([fs.readFile(file), fs.stat(file)]);
+  return {
+    hash: createHash("sha256").update(raw).digest("hex"),
+    mtimeMs: stat.mtimeMs,
+    size: stat.size
+  };
+}
+
+async function withBriefReadOnlyGuard(cwd: string, task: () => Promise<void>): Promise<void> {
+  const before = await snapshotBrief(cwd);
+  await task();
+  const after = await snapshotBrief(cwd);
+  if (!before) return;
+  if (!after) {
+    throw new UserError("Input file `brief.md` was removed during execution. Input files are read-only.");
+  }
+  if (before.hash !== after.hash || before.size !== after.size || before.mtimeMs !== after.mtimeMs) {
+    throw new UserError("Input file `brief.md` was modified during execution. Input files are read-only.");
+  }
+}
+
+export async function runCli(options: RunOptions = {}): Promise<number> {
+  const cwd = options.cwd ?? process.cwd();
+  const argv = options.argv ?? process.argv;
+  const out = options.log ?? console.log;
+  const err = options.error ?? console.error;
+  const forced = options.forcedCommand ? mapForcedCommand(options.forcedCommand) : undefined;
+
+  const program = new Command();
+  const version = await readCliVersion(cwd);
+  program
+    .name("prodo")
+    .description("CLI-first, prompt-powered product artifact kit")
+    .version(`prodo ${version}`, "-v, --version", "Show Prodo version")
+    .showHelpAfterError();
+  const artifactTypes = await listArtifactTypes(cwd);
+
+  program
+    .command("init [target]")
+    .option("--ai <name>", "agent integration: codex | gemini-cli | claude-cli")
+    .option("--lang <code>", "document language (e.g. en, tr)")
+    .option("--author <name>", "document author name")
+    .option("--preset <name>", "preset to install during initialization")
+    .action(async (target, opts) => {
+      const projectRoot = path.resolve(cwd, target ?? ".");
+      const selected = await gatherInitSelections({
+        projectRoot,
+        aiInput: opts.ai,
+        langInput: opts.lang,
+        authorInput: opts.author
+      });
+      const selectedAi = selected.ai as SupportedAi | undefined;
+
+      if (selected.interactive) {
+        const clack = (await dynamicImport("@clack/prompts")) as typeof import("@clack/prompts");
+        const s = clack.spinner();
+        s.start("Scaffolding Prodo workspace...");
+        const result = await runInit(projectRoot, {
+          ai: selectedAi,
+          lang: selected.lang,
+          author: selected.author,
+          preset: opts.preset,
+          script: selected.script
+        });
+        s.stop("Scaffold complete.");
+        await finishInitInteractive({
+          projectRoot,
+          settingsPath: result.settingsPath,
+          ai: selectedAi,
+          lang: selected.lang,
+          author: selected.author
+        });
+        return;
+      }
+
+      const result = await runInit(projectRoot, {
+        ai: selectedAi,
+        lang: selected.lang,
+        author: selected.author,
+        preset: opts.preset,
+        script: selected.script
+      });
+      out(`Initialized Prodo scaffold at ${path.join(projectRoot, ".prodo")}`);
+      if (selectedAi) {
+        out(`Agent command set installed for ${selectedAi}.`);
+        out(`Installed ${result.installedAgentFiles.length} command files.`);
+        out("Agent workflow: edit brief.md, then run slash commands in your agent.");
+      } else {
+        out("No agent selected. Use `prodo generate` for end-to-end generation.");
+      }
+      out(`Settings file: ${result.settingsPath}`);
+      out(`Author: ${selected.author}`);
+      out("Next: edit brief.md.");
+    });
+
+  program
+    .command("generate")
+    .description("Run end-to-end pipeline: normalize -> generate artifacts -> validate")
+    .option("--agent <name>", "agent profile: codex | gemini-cli | claude-cli")
+    .option("--strict", "treat validation warnings as errors")
+    .option("--report <path>", "validation report output path")
+    .option("--dry-run", "show what would be generated without writing files")
+    .action(async (opts: { agent?: string; strict?: boolean; report?: string; dryRun?: boolean }) => {
+      if (opts.agent) resolveAgent(opts.agent);
+      if (opts.dryRun) {
+        out("[Dry Run] Pipeline would execute:");
+        out(`  1. Normalize brief.md`);
+        for (const type of artifactTypes) {
+          out(`  2. Generate ${type}`);
+        }
+        out(`  3. Validate all artifacts`);
+        out(`\nArtifact types: ${artifactTypes.join(", ")}`);
+        return;
+      }
+      await withBriefReadOnlyGuard(cwd, async () => {
+        await runHookPhase(cwd, "before_normalize", out);
+        const normalizedPath = await runNormalize({ cwd });
+        out(`Normalized brief written to: ${normalizedPath}`);
+        await runHookPhase(cwd, "after_normalize", out);
+
+        for (const type of artifactTypes) {
+          await runArtifactCommand(type, { from: normalizedPath, agent: opts.agent }, cwd, out, {
+            suggestValidate: false
+          });
+        }
+
+        await runHookPhase(cwd, "before_validate", out);
+        const result = await runValidate(cwd, {
+          strict: Boolean(opts.strict),
+          report: opts.report
+        });
+        out(`Validation report written to: ${result.reportPath}`);
+        if (!result.pass) {
+          throw new UserError("Validation failed. Review report and fix issues.");
+        }
+        out("Generation pipeline completed. Validation passed.");
+        await runHookPhase(cwd, "after_validate", out);
+      });
+    });
+
+  program
+    .command("fix", { hidden: true })
+    .description("Advanced: auto-regenerate affected artifacts from validation findings")
+    .option("--agent <name>", "agent profile: codex | gemini-cli | claude-cli")
+    .option("--strict", "treat validation warnings as errors")
+    .option("--report <path>", "validation report output path")
+    .option("--dry-run", "show fix proposal without applying changes")
+    .action(async (opts: { agent?: string; strict?: boolean; report?: string; dryRun?: boolean }) => {
+      if (opts.agent) resolveAgent(opts.agent);
+      await withBriefReadOnlyGuard(cwd, async () => {
+        const fixOpts = {
+          cwd,
+          agent: opts.agent,
+          strict: Boolean(opts.strict),
+          report: opts.report,
+          dryRun: Boolean(opts.dryRun),
+          log: out
+        };
+
+        if (opts.dryRun) {
+          const result = await runFix(fixOpts);
+          out(`Validation report: ${result.reportPath}`);
+          if (result.proposal.targets.length > 0) {
+            await displayFixProposal(result.proposal, out);
+          }
+          return;
+        }
+
+        const proposal = await buildFixProposal(fixOpts);
+        out(`Validation report: ${proposal.initialReport.reportPath}`);
+
+        if (proposal.targets.length === 0) {
+          out("No blocking issues found. Nothing to fix.");
+          return;
+        }
+
+        await displayFixProposal(proposal, out);
+
+        const confirmed = await confirmFixExecution(proposal);
+        if (!confirmed) {
+          out("Fix cancelled by user.");
+          return;
+        }
+
+        out(`Regenerating impacted artifacts: ${proposal.targets.join(", ")}`);
+        const result = await applyFix(cwd, proposal, fixOpts);
+
+        await displayFixResult(result, out);
+
+        if (!result.finalPass) {
+          throw new UserError("Fix completed but validation is still failing. Review report and retry.");
+        }
+      });
+    });
+
+  program
+    .command("normalize", { hidden: true })
+    .description("Advanced: normalize brief without full pipeline")
+    .option("--brief <path>", "path to start brief markdown")
+    .option("--out <path>", "output normalized brief json path")
+    .option("--agent <name>", "agent profile: codex | gemini-cli | claude-cli")
+    .option("-i, --interactive", "interactively clarify low-confidence fields")
+    .option("--dry-run", "show what would be normalized without writing")
+    .action(async (opts: { brief?: string; out?: string; agent?: string; interactive?: boolean; dryRun?: boolean }) => {
+      if (opts.agent) resolveAgent(opts.agent);
+      if (opts.dryRun) {
+        const briefFile = opts.brief ?? "brief.md";
+        out(`[Dry Run] Would normalize: ${briefFile}`);
+        out(`[Dry Run] Output would be written to: .prodo/briefs/normalized-brief.json`);
+        return;
+      }
+      await withBriefReadOnlyGuard(cwd, async () => {
+        await runHookPhase(cwd, "before_normalize", out);
+        const outPath = opts.interactive
+          ? await runInteractiveNormalize({ cwd, brief: opts.brief, out: opts.out, log: out })
+          : await runNormalize({ cwd, brief: opts.brief, out: opts.out });
+        out(`Normalized brief written to: ${outPath}`);
+        await runHookPhase(cwd, "after_normalize", out);
+      });
+    });
+
+  program
+    .command("doctor")
+    .alias("check")
+    .description("Check local environment and toolchain readiness")
+    .action(async () => {
+      await runDoctor(cwd, out);
+    });
+
+  program
+    .command("clean")
+    .description("Remove all generated artifacts, keep brief.md and config")
+    .option("--dry-run", "show what would be removed without deleting")
+    .action(async (opts: { dryRun?: boolean }) => {
+      const result = await runClean({
+        cwd,
+        dryRun: Boolean(opts.dryRun),
+        log: out
+      });
+      if (result.removedPaths.length === 0) {
+        out("Nothing to clean.");
+      } else if (!opts.dryRun) {
+        out(`Cleaned ${result.removedPaths.length} path(s). Project is ready for a fresh run.`);
+      }
+    });
+
+  for (const type of artifactTypes) {
+    program
+      .command(type, { hidden: true })
+      .description(`Advanced: generate only ${type} artifact`)
+      .option("--from <path>", "path to normalized-brief.json")
+      .option("--out <path>", "output file path")
+      .option("--agent <name>", "agent profile: codex | gemini-cli | claude-cli")
+      .action(async (opts: { from?: string; out?: string; agent?: string }) => {
+        await withBriefReadOnlyGuard(cwd, async () => {
+          await runArtifactCommand(type, opts, cwd, out);
+        });
+      });
+  }
+
+  program
+    .command("agent-commands", { hidden: true })
+    .requiredOption("--agent <name>", "agent profile: codex | gemini-cli | claude-cli")
+    .action(async (opts: { agent: string }) => {
+      const agent = resolveAgent(opts.agent);
+      if (!agent) throw new UserError("Agent is required.");
+      const set = await loadAgentCommandSet(cwd, agent);
+      out(`Agent: ${set.agent}`);
+      if (set.description) out(`Description: ${set.description}`);
+      out("");
+      out("Recommended sequence:");
+      for (const item of set.recommended_sequence ?? []) {
+        out(`- ${item.command}: ${item.purpose}`);
+      }
+      if (set.artifact_shortcuts) {
+        out("");
+        out("Artifact shortcuts:");
+        for (const [key, command] of Object.entries(set.artifact_shortcuts)) {
+          out(`- ${key}: ${command}`);
+        }
+      }
+    });
+
+  program
+    .command("validate", { hidden: true })
+    .description("Advanced: run validation only")
+    .option("--strict", "treat warnings as errors")
+    .option("--report <path>", "report output path")
+    .option("--agent <name>", "agent profile: codex | gemini-cli | claude-cli")
+    .action(async (opts: { strict?: boolean; report?: string; agent?: string }) => {
+      if (opts.agent) resolveAgent(opts.agent);
+      await withBriefReadOnlyGuard(cwd, async () => {
+        await runHookPhase(cwd, "before_validate", out);
+        const result = await runValidate(cwd, {
+          strict: Boolean(opts.strict),
+          report: opts.report
+        });
+        out(`Validation report written to: ${result.reportPath}`);
+        if (!result.pass) {
+          throw new UserError("Validation failed. Review report and fix issues.");
+        }
+        out("Validation passed.");
+        await runHookPhase(cwd, "after_validate", out);
+      });
+    });
+
+  program
+    .command("skills", { hidden: true })
+    .description("Advanced: manage and run skills")
+    .argument("[action]", "list or run", "list")
+    .argument("[name]", "skill name (for run)")
+    .option("--input <json>", "JSON input for skill execution")
+    .action(async (action: string, name: string | undefined, opts: { input?: string }) => {
+      const { getGlobalSkillEngine } = await import("../skills/engine");
+      const engine = getGlobalSkillEngine();
+
+      if (action === "list") {
+        const manifests = engine.listSkills();
+        if (manifests.length === 0) {
+          out("No skills registered.");
+          return;
+        }
+        out("Available skills:\n");
+        for (const m of manifests) {
+          out(`  ${m.name.padEnd(25)} [${m.category}] ${m.description}`);
+        }
+        return;
+      }
+
+      if (action === "run") {
+        if (!name) throw new UserError("Skill name is required. Usage: prodo skills run <name>");
+        const inputs = opts.input ? JSON.parse(opts.input) as Record<string, unknown> : {};
+        inputs.cwd = inputs.cwd ?? cwd;
+        const result = await engine.execute(name, { cwd, log: out }, inputs);
+        out(`\nSkill "${name}" completed.`);
+        out(JSON.stringify(result, null, 2));
+        return;
+      }
+
+      throw new UserError(`Unknown skills action: "${action}". Use: list or run`);
+    });
+
+  try {
+    if (forced) {
+      if (forced === "init") {
+        await program.parseAsync(["node", "prodo", "init", ...argv.slice(2)]);
+      } else if (forced === "normalize") {
+        await program.parseAsync(["node", "prodo", "normalize", ...argv.slice(2)]);
+      } else if (forced === "validate") {
+        await program.parseAsync(["node", "prodo", "validate", ...argv.slice(2)]);
+      } else if (forced === "fix") {
+        await program.parseAsync(["node", "prodo", "fix", ...argv.slice(2)]);
+      } else {
+        await program.parseAsync(["node", "prodo", forced, ...argv.slice(2)]);
+      }
+    } else {
+      await program.parseAsync(argv);
+    }
+    return 0;
+  } catch (error) {
+    if (error instanceof UserError) {
+      err(error.message);
+      return 1;
+    }
+    const unknown = error as Error;
+    err(unknown.message);
+    return 1;
+  }
+}
+
+if (require.main === module) {
+  runCli().then((code) => {
+    process.exitCode = code;
+  });
+}
+