chatroom-cli 1.39.3 → 1.40.0

package/dist/index.js

@@ -71238,6 +71238,286 @@ function readFileContent(filePath, optionName) {
 }
 var init_file_content = () => {};
 
+// src/commands/workflow/templates/code-review.ts
+function getCodeReviewTemplate(role) {
+  return {
+    key: "code-review",
+    steps: [
+      {
+        stepKey: "pillar-1-simplification",
+        description: "Simplification: functions/classes, dead code, monolithic patterns",
+        dependsOn: [],
+        order: 1,
+        assigneeRole: role,
+        specification: {
+          goal: `## Pillar 1 — Simplification (Highest Priority)
+
+AI generates more code than needed. The average developer checked in 75% more code in 2025
+than in 2022 — volume that the team now has to maintain. AI agents never suggest refactoring,
+so complexity accumulates silently.
+
+Look for:
+- Functions over ~40 lines or classes over ~200 lines with no clear single responsibility
+- Monolithic architecture: 40–50% of AI code defaults to tightly-coupled structures that
+  reverse a decade of modular progress (OX Security, 2025)
+- Phantom Bugs: logic handling highly improbable edge cases, adding complexity with no benefit
+  (found in 20–30% of AI code)
+- Dead code, unused imports, or leftover scaffolding from generation
+- PRs that touch services, libraries, infrastructure, and tests in a single change
+
+Ask: Is there a simpler path to the same outcome? Would a new team member understand this
+without help? Can this function be split without losing meaning?
+
+Action: Flag any function over ~40 lines. Require decomposition or written justification
+before merge.`,
+          requirements: REVIEW_REQUIREMENTS
+        }
+      },
+      {
+        stepKey: "pillar-2-type-drift",
+        description: "Type Drift: any usage, unsafe assertions, null checks, boundary drift",
+        dependsOn: ["pillar-1-simplification"],
+        order: 2,
+        assigneeRole: role,
+        specification: {
+          goal: `## Pillar 2 — Type Drift (High Priority)
+
+AI generates code in isolation from the broader type system. Type inconsistencies introduced
+here surface later as null pointer exceptions, silent failures, and cross-service contract
+mismatches. A University of Naples study of 500k+ samples (Aug 2025) confirmed these are the
+most reliably catchable issues via static analysis — meaning they should never reach review.
+
+Look for:
+- \`any\` usage: defeats TypeScript entirely; AI reaches for it when uncertain
+- Unsafe \`as\` assertions: suppresses the compiler rather than satisfying it
+- Missing \`strictNullChecks\` compliance — null/undefined slipping through unchecked
+- Boundary drift: API endpoints returning different shapes across execution paths
+- Configuration drift: divergent \`tsconfig\` settings or ESLint rules added incrementally
+- Leaked internals: AI exposing implementation details that should be encapsulated
+
+Ask: Are all return types and parameters explicitly typed? Are API contract types shared and
+stable? Does this match existing type conventions, or introduce a new pattern?
+
+Action: \`strict: true\` in \`tsconfig.json\` is a hard gate. Any PR disabling strict settings
+requires explicit team sign-off.`,
+          requirements: REVIEW_REQUIREMENTS
+        }
+      },
+      {
+        stepKey: "pillar-3-duplication",
+        description: "Duplication: shared abstractions, copy-paste, vanilla rebuilds",
+        dependsOn: ["pillar-2-type-drift"],
+        order: 3,
+        assigneeRole: role,
+        specification: {
+          goal: `## Pillar 3 — Duplication (High Priority)
+
+GitClear's 2025 analysis of 211 million changed lines found duplicated code blocks grew 8x
+in one year, and copy/pasted code surpassed refactored code for the first time in history.
+OX Security found "Bugs Déjà-Vu" in 70–80% of AI codebases: identical bugs recurring in
+multiple locations because logic was never abstracted.
+
+Look for:
+- Duplicate utility logic across files — each instance works, but no shared abstraction exists
+- "Vanilla Style" rebuilds: AI reconstructing functionality from scratch instead of using
+  existing libraries, SDKs, or internal utilities (40–50% of AI code, OX Security)
+- Duplicated validation, auth checks, and input sanitization
+- Copy-pasted try/catch blocks instead of centralised error handling
+- New dependencies that the existing codebase already covers
+
+Ask: Does an equivalent utility already exist? Is this logic duplicated anywhere else?
+Use AST-level search, not just grep.
+
+Action: Require a codebase search before approving any new utility function. Run SonarQube,
+CodeClimate, or Codacy as a PR gate for duplication thresholds.`,
+          requirements: REVIEW_REQUIREMENTS
+        }
+      },
+      {
+        stepKey: "pillar-4-design-patterns",
+        description: "Design Patterns: SOLID compliance, pattern violations, context blindness",
+        dependsOn: ["pillar-3-duplication"],
+        order: 4,
+        assigneeRole: role,
+        specification: {
+          goal: `## Pillar 4 — Design Pattern Compliance (Standard Priority)
+
+AI implements the prompt. It does not check the ADR log, existing service patterns, or
+architectural conventions. OX Security identified "By-The-Book Fixation" in 80–90% of AI
+code: rigid application of generic best practices, missing the better solution for this
+specific context.
+
+Look for:
+- Pattern violations: raw \`fetch\` where an API service layer exists; new state patterns where
+  a store convention is already established
+- SOLID violations:
+  - Single Responsibility: see Pillar 1
+  - Open/Closed: code requiring modification rather than extension to add new behaviour
+  - Dependency Inversion: hard-coded dependencies instead of injection
+- No refactoring suggestions — AI never proposes that a feature should trigger a refactor
+- Context blindness: ignoring ADRs, versioning strategy, or team-level conventions
+
+Ask: Does this follow the patterns in this area of the codebase? Can this be extended without
+modification? Are dependencies injected?
+
+Action: Maintain a living \`ARCHITECTURE.md\` or ADR log. Pass it as context to AI tools before
+generation, not after.`,
+          requirements: REVIEW_REQUIREMENTS
+        }
+      },
+      {
+        stepKey: "pillar-5-security",
+        description: "Security: OWASP Top 10, auth checks, secrets, env awareness",
+        dependsOn: ["pillar-4-design-patterns"],
+        order: 5,
+        assigneeRole: role,
+        specification: {
+          goal: `## Pillar 5 — Security (Standard Priority)
+
+45% of AI-generated code samples introduced OWASP Top 10 vulnerabilities in Veracode's 2025
+test of 100+ LLMs. CodeRabbit found AI code is 2.74x more likely to introduce XSS, 1.88x more
+likely to mishandle passwords, 1.82x more likely to implement insecure deserialization.
+OX Security calls this "insecure by dumbness" — not malicious, but structurally blind to
+security requirements that were not in the prompt.
+
+Look for:
+- OWASP Top 10: injection (SQL, XSS, command), broken auth, insecure deserialization
+- "Worked on My Machine" syndrome (60–70% of AI code): missing environment awareness,
+  hard-coded localhost references, assumed-present secrets
+- Missing auth/ownership checks — AI assumes the happy path; role gates rarely appear unprompted
+- Secrets, tokens, or credentials in source code
+- Deprecated API patterns from the model's training data
+- Excessive I/O (~8x human rate) and concurrency misuse (~2x human rate)
+
+Ask: Are all inputs validated before use? Are auth checks present at every trust boundary?
+Does this code behave differently across environments?
+
+Action: Run SAST tooling (Snyk, Semgrep, SonarQube) on every AI-generated PR as a CI gate.
+Manual review alone cannot keep pace with AI generation velocity.`,
+          requirements: REVIEW_REQUIREMENTS
+        }
+      },
+      {
+        stepKey: "pillar-6-test-quality",
+        description: "Test Quality: edge cases, coverage theater, integration tests",
+        dependsOn: ["pillar-5-security"],
+        order: 6,
+        assigneeRole: role,
+        specification: {
+          goal: `## Pillar 6 — Test Quality (Standard Priority)
+
+AI generates tests that look comprehensive but frequently are not. OX Security found "Fake
+Test Coverage" in 40–50% of AI codebases: inflated metrics, low signal. The failure mode is
+precise: AI tests its own assumptions, not the developer's intent. University of Naples (2025)
+confirmed that even code passing all functional benchmarks averaged 1.45–1.77 static issues
+per task.
+
+Look for:
+- Tests that verify AI output is stable, not that it is correct by domain rules
+- Missing edge cases: empty inputs, boundary values, null handling, race conditions,
+  business-rule violations
+- Coverage theater: tests that assert no exception was thrown and nothing more
+- No integration or contract tests — AI generates unit tests almost exclusively
+- Tests coupled to implementation: must be updated every time the code changes
+
+Ask: Do these tests verify the domain intent? Would they catch a logic regression in the
+next PR? Are cross-service interactions tested?
+
+Action: Test review is a separate pass from code review. Use mutation testing (Stryker,
+PITest) to validate whether tests catch real defects — coverage percentage is not sufficient.`,
+          requirements: REVIEW_REQUIREMENTS
+        }
+      },
+      {
+        stepKey: "pillar-7-ownership",
+        description: "Ownership & Observability: named owner, logging, error handling, health checks",
+        dependsOn: ["pillar-6-test-quality"],
+        order: 7,
+        assigneeRole: role,
+        specification: {
+          goal: `## Pillar 7 — Ownership, Observability, Deployment (Standard Priority)
+
+The most consistently missed category. ISACA's 2026 incident review found that the biggest
+AI failures of 2025 were organisational, not technical: weak controls and unclear ownership.
+Bright Security (2026) identified unclear ownership as one of the most dangerous patterns:
+code that works well enough that no one feels responsible for it. IBM's 2025 breach data found
+shadow AI added $670,000 average to breach costs.
+
+Look for:
+- No named human owner — every AI-generated module needs someone who can explain it in a
+  postmortem without reading it fresh
+- Missing structured logs, metrics emission, or tracing hooks
+- Swallowed exceptions: retry logic with no alerting; fallback paths invisible to operators
+- No environment-specific configuration, secrets management, or health check endpoints
+- Model versioning chaos: inconsistencies from team members using different AI tool versions
+- Shadow AI: code with no record of which tool or prompt produced it
+
+Ask: Who owns this? Will production failures surface to operators? Does this behave
+differently across environments?
+
+Action: Require every AI-generated module to have a named owner in CODEOWNERS. Add logging
+requirements to the PR template. Establish a team AI governance policy that tracks tools
+and model versions — treat this like any other dependency.`,
+          requirements: REVIEW_REQUIREMENTS
+        }
+      },
+      {
+        stepKey: "pillar-8-dead-code",
+        description: "Dead Code: unreachable code, unused imports, orphaned config",
+        dependsOn: ["pillar-7-ownership"],
+        order: 8,
+        assigneeRole: role,
+        specification: {
+          goal: `## Pillar 8 — Dead Code Elimination (Standard Priority)
+
+AI generates code speculatively. It adds helper functions "just in case", creates abstractions
+for paths that never materialise, and leaves behind scaffolding from earlier generation
+attempts. GitClear's 2025 data showed a 75% increase in total code volume, much of it never
+executed. Dead code is not harmless — it misleads future readers, inflates bundle sizes, creates
+false positive search results, and adds surface area for bugs and security vulnerabilities
+in code that serves no purpose.
+
+Look for:
+- Unreachable code: functions, methods, or branches that no call site ever invokes
+- Unused imports: modules imported but never referenced — common in AI-generated files
+- Commented-out code: old implementations left inline instead of being removed; version
+  control already preserves history
+- Feature flags that are permanently off: conditional paths that were never enabled or have
+  been superseded but not cleaned up
+- Orphaned configuration: environment variables, constants, or config entries with no consumer
+- Stale types and interfaces: type definitions for data shapes that no longer exist in the
+  system
+- Unused dependencies: packages listed in \`package.json\` (or equivalent) that no source file
+  imports
+- Vestigial error handling: catch blocks or fallback paths for conditions that can no longer
+  occur due to upstream changes
+- Test helpers and fixtures for tests that have been deleted
+
+Ask: Is every function called? Is every import used? Is every dependency exercised?
+Would removing this code change any observable behaviour?
+
+Action: Run tree-shaking analysis and dead code detection tools (e.g., \`ts-prune\`,
+\`knip\`, \`depcheck\`, IDE unused-symbol highlighting) as part of CI. Treat dead code
+the same as duplication — it compounds over time and must be actively managed.`,
+          requirements: REVIEW_REQUIREMENTS
+        }
+      }
+    ]
+  };
+}
+var REVIEW_REQUIREMENTS = "Review the code against this pillar. Mark this step complete when you have finished reviewing and noting all findings for this pillar.";
+
+// src/commands/workflow/templates/index.ts
+function getTemplate(name, role) {
+  switch (name.toLowerCase()) {
+    case "code-review":
+      return getCodeReviewTemplate(role);
+    default:
+      return null;
+  }
+}
+var init_templates = () => {};
+
 // src/commands/workflow/index.ts
 var exports_workflow = {};
 __export(exports_workflow, {
@@ -71247,6 +71527,7 @@ __export(exports_workflow, {
   getWorkflowStatus: () => getWorkflowStatus,
   exitWorkflow: () => exitWorkflow,
   executeWorkflow: () => executeWorkflow,
+  createWorkflowFromTemplate: () => createWorkflowFromTemplate,
   createWorkflow: () => createWorkflow,
   completeStep: () => completeStep
 });
@@ -71599,17 +71880,33 @@ async function completeStep(chatroomId, options, deps) {
   const sessionId = await requireAuth3(d);
   validateChatroomId2(chatroomId);
   try {
-    await d.backend.mutation(api.workflows.completeStep, {
+    const result = await d.backend.mutation(api.workflows.completeStep, {
       sessionId,
       chatroomId,
       workflowKey: options.workflowKey,
       stepKey: options.stepKey
     });
     console.log("");
-    console.log("✅ Step completed");
+    console.log("✅ Step complete!");
     console.log(`   Workflow: ${options.workflowKey}`);
     console.log(`   Step: ${options.stepKey}`);
     console.log("");
+    if (result.workflowComplete) {
+      console.log("\uD83C\uDF89 All steps complete — the workflow is done.");
+      console.log("");
+    } else if (result.nextStep) {
+      if (result.nextStep.assigneeRole === options.role) {
+        console.log(`⏭️  Next step unlocked: "${result.nextStep.description}"`);
+        console.log("   Run this to view it:");
+        console.log(`   chatroom workflow step-view --chatroom-id=${chatroomId} --role=${options.role} --workflow-key=${options.workflowKey} --step-key=${result.nextStep.stepKey}`);
+        console.log("");
+      } else {
+        const assignee = result.nextStep.assigneeRole ?? "unknown";
+        console.log(`⏭️  Next step unlocked: "${result.nextStep.description}"`);
+        console.log(`   Assigned to: ${assignee}`);
+        console.log("");
+      }
+    }
   } catch (error) {
     console.error(`❌ Failed to complete step: ${getErrorMessage(error)}`);
     process.exit(1);
@@ -71732,12 +72029,77 @@ async function viewStep(chatroomId, options, deps) {
     return;
   }
 }
+async function createWorkflowFromTemplate(chatroomId, options, deps) {
+  const d = deps ?? await createDefaultDeps12();
+  const sessionId = await requireAuth3(d);
+  validateChatroomId2(chatroomId);
+  const template = getTemplate(options.template, options.role);
+  if (!template) {
+    console.error(`❌ Unknown template: "${options.template}"`);
+    console.error("   Available templates: code-review");
+    process.exit(1);
+    return;
+  }
+  const workflowKey = `${template.key}-${Date.now()}`;
+  try {
+    const stepsForCreate = template.steps.map((s) => ({
+      stepKey: s.stepKey,
+      description: s.description,
+      dependsOn: s.dependsOn,
+      order: s.order
+    }));
+    const createResult = await d.backend.mutation(api.workflows.createWorkflow, {
+      sessionId,
+      chatroomId,
+      workflowKey,
+      steps: stepsForCreate,
+      createdBy: options.role
+    });
+    console.log("");
+    console.log(`✅ Workflow created from template "${options.template}"`);
+    console.log(`   Key: ${workflowKey}`);
+    console.log(`   Steps: ${template.steps.length} (one pillar revealed at a time)`);
+    for (const step4 of template.steps) {
+      if (step4.specification) {
+        await d.backend.mutation(api.workflows.specifyStep, {
+          sessionId,
+          chatroomId,
+          workflowKey,
+          stepKey: step4.stepKey,
+          assigneeRole: step4.assigneeRole,
+          goal: step4.specification.goal,
+          requirements: step4.specification.requirements,
+          warnings: step4.specification.warnings
+        });
+      }
+    }
+    console.log(`   Specifications: ${template.steps.length} steps specified`);
+    await d.backend.mutation(api.workflows.executeWorkflow, {
+      sessionId,
+      chatroomId,
+      workflowKey
+    });
+    console.log(`   Status: active`);
+    console.log("");
+    console.log("Next steps:");
+    const firstStep = template.steps[0];
+    console.log(`   1. View first step: chatroom workflow step-view --chatroom-id=${chatroomId} --role=${options.role} --workflow-key=${workflowKey} --step-key=${firstStep.stepKey}`);
+    console.log(`   2. When done reviewing: chatroom workflow step-complete --chatroom-id=${chatroomId} --role=${options.role} --workflow-key=${workflowKey} --step-key=${firstStep.stepKey}`);
+    console.log("   3. Proceed to next pillar as each step unlocks");
+    console.log("");
+  } catch (error) {
+    console.error(`❌ Failed to start workflow: ${getErrorMessage(error)}`);
+    process.exit(1);
+    return;
+  }
+}
 var VALID_CHATROOM_SKILLS;
 var init_workflow = __esm(() => {
   init_api3();
   init_storage();
   init_client2();
   init_convex_error();
+  init_templates();
   VALID_CHATROOM_SKILLS = ["backlog", "software-engineering", "code-review", "workflow"];
 });
 
@@ -79084,6 +79446,14 @@ backlogCommand.command("import").description("Import backlog items from a JSON e
   await importBacklog2(options.chatroomId, { role: options.role, path: options.path });
 });
 var workflowCommand = program2.command("workflow").description("Manage structured workflows");
+workflowCommand.command("create-from-template").description("Create a workflow from a built-in template with all steps pre-filled").requiredOption("--chatroom-id <id>", "Chatroom identifier").requiredOption("--role <role>", "Your role").requiredOption("--template <name>", "Template name (e.g. code-review)").action(async (options) => {
+  await maybeRequireAuth();
+  const { createWorkflowFromTemplate: createWorkflowFromTemplate2 } = await Promise.resolve().then(() => (init_workflow(), exports_workflow));
+  await createWorkflowFromTemplate2(options.chatroomId, {
+    role: options.role,
+    template: options.template
+  });
+});
 workflowCommand.command("create").description("Create a new workflow with steps (reads JSON from stdin)").requiredOption("--chatroom-id <id>", "Chatroom identifier").requiredOption("--role <role>", "Your role (creator)").requiredOption("--workflow-key <key>", "Unique workflow key").action(async (options) => {
   await maybeRequireAuth();
   const stdinContent = await readStdin();
@@ -79339,5 +79709,5 @@ program2.hook("preAction", async (_thisCommand, actionCommand) => {
 });
 program2.parse();
 
-//# debugId=F3CC073413306D7A64756E2164756E21
+//# debugId=CC4268FCAB22FC9D64756E2164756E21
 //# sourceMappingURL=index.js.map