diff --git a/.archon/workflows/defaults/archon-interactive-prd.yaml b/.archon/workflows/defaults/archon-interactive-prd.yaml
index ccb08cb411..79024445d6 100644
--- a/.archon/workflows/defaults/archon-interactive-prd.yaml
+++ b/.archon/workflows/defaults/archon-interactive-prd.yaml
@@ -52,6 +52,7 @@ nodes:
   - id: foundation-gate
     approval:
       message: "Answer the foundation questions above. Your answers will guide the research phase."
+      capture_response: true
     depends_on: [initiate]
 
   # ═══════════════════════════════════════════════════════════════
@@ -106,6 +107,7 @@ nodes:
   - id: deepdive-gate
     approval:
       message: "Answer the deep dive questions above (vision, primary user, JTBD, constraints). Add any adjustments from the research."
+      capture_response: true
     depends_on: [research]
 
   # ═══════════════════════════════════════════════════════════════
@@ -172,6 +174,7 @@ nodes:
   - id: scope-gate
     approval:
       message: "Answer the scope questions above (MVP, must-haves, hypothesis, exclusions). This is the final input before PRD generation."
+      capture_response: true
     depends_on: [technical]
 
   # ═══════════════════════════════════════════════════════════════
@@ -188,11 +191,11 @@ nodes:
       **Deep dive answers**: $deepdive-gate.output
       **Scope answers**: $scope-gate.output
 
-      Generate a complete PRD file at `.claude/PRPs/prds/{kebab-case-name}.prd.md`.
+      Generate a complete PRD file at `$ARTIFACTS_DIR/prds/{kebab-case-name}.prd.md`.
 
       First create the directory:
       ```bash
-      mkdir -p .claude/PRPs/prds
+      mkdir -p $ARTIFACTS_DIR/prds
       ```
 
       **First principles rule**: Before writing the Technical Approach section, READ the
@@ -243,9 +246,9 @@ nodes:
       Read the PRD file that was just generated. The generate node output the file path:
       $generate.output
 
-      Find the PRD file — check `.claude/PRPs/prds/` for the most recently created `.prd.md` file:
+      Find the PRD file — check `$ARTIFACTS_DIR/prds/` for the most recently created `.prd.md` file:
       ```bash
-      ls -t .claude/PRPs/prds/*.prd.md | head -1
+      ls -t $ARTIFACTS_DIR/prds/*.prd.md | head -1
       ```
 
       Read the entire PRD, then verify EVERY technical claim against the actual codebase:
diff --git a/CLAUDE.md b/CLAUDE.md
index 4448b38eff..012498439f 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -672,7 +672,7 @@ async function createSession(conversationId: string, codebaseId: string) {
 2. **Workflows** (YAML-based):
    - Stored in `.archon/workflows/` (searched recursively)
    - Multi-step AI execution chains, discovered at runtime
-   - **`nodes:` (DAG format)**: Nodes with explicit `depends_on` edges; independent nodes in the same topological layer run concurrently. Node types: `command:` (named command file), `prompt:` (inline prompt), `bash:` (shell script, stdout captured as `$nodeId.output`, no AI), `loop:` (iterative AI prompt until completion signal), `script:` (inline TypeScript/Python or named script from `.archon/scripts/`, runs via `bun` or `uv`, stdout captured as `$nodeId.output`, no AI, supports `deps:` for dependency installation and `timeout:` in ms, requires `runtime: bun` or `runtime: uv`) . Supports `when:` conditions, `trigger_rule` join semantics, `$nodeId.output` substitution, `output_format` for structured JSON output (Claude and Codex), `allowed_tools`/`denied_tools` for per-node tool restrictions (Claude only), `hooks` for per-node SDK hook callbacks (Claude only), `mcp` for per-node MCP server config files (Claude only, env vars expanded at execution time), and `skills` for per-node skill preloading via AgentDefinition wrapping (Claude only), and `effort`/`thinking`/`maxBudgetUsd`/`systemPrompt`/`fallbackModel`/`betas`/`sandbox` for Claude SDK advanced options (Claude only, also settable at workflow level)
+   - **`nodes:` (DAG format)**: Nodes with explicit `depends_on` edges; independent nodes in the same topological layer run concurrently. Node types: `command:` (named command file), `prompt:` (inline prompt), `bash:` (shell script, stdout captured as `$nodeId.output`, no AI), `loop:` (iterative AI prompt until completion signal), `approval:` (human gate; pauses until user approves or rejects; `capture_response: true` stores the user's comment as `$<node-id>.output` for downstream nodes, default false), `script:` (inline TypeScript/Python or named script from `.archon/scripts/`, runs via `bun` or `uv`, stdout captured as `$nodeId.output`, no AI, supports `deps:` for dependency installation and `timeout:` in ms, requires `runtime: bun` or `runtime: uv`) . Supports `when:` conditions, `trigger_rule` join semantics, `$nodeId.output` substitution, `output_format` for structured JSON output (Claude and Codex), `allowed_tools`/`denied_tools` for per-node tool restrictions (Claude only), `hooks` for per-node SDK hook callbacks (Claude only), `mcp` for per-node MCP server config files (Claude only, env vars expanded at execution time), and `skills` for per-node skill preloading via AgentDefinition wrapping (Claude only), and `effort`/`thinking`/`maxBudgetUsd`/`systemPrompt`/`fallbackModel`/`betas`/`sandbox` for Claude SDK advanced options (Claude only, also settable at workflow level)
    - Provider inherited from `.archon/config.yaml` unless explicitly set; per-node `provider` and `model` overrides supported
    - Model and options can be set per workflow or inherited from config defaults
    - `interactive: true` at the workflow level forces foreground execution on web (required for approval-gate workflows in the web UI)
diff --git a/packages/cli/src/commands/workflow.test.ts b/packages/cli/src/commands/workflow.test.ts
index 0a050cf26f..7f13f8d83f 100644
--- a/packages/cli/src/commands/workflow.test.ts
+++ b/packages/cli/src/commands/workflow.test.ts
@@ -107,6 +107,7 @@ mock.module('@archon/core/db/conversations', () => ({
   getOrCreateConversation: mock(() =>
     Promise.resolve({ id: 'conv-123', platform_type: 'cli', platform_conversation_id: 'cli-123' })
   ),
+  getConversationById: mock(() => Promise.resolve(null)),
   updateConversation: mock(() => Promise.resolve()),
 }));
 
@@ -1381,6 +1382,58 @@ describe('workflowApproveCommand', () => {
 
     expect(codebaseDb.getCodebase).toHaveBeenCalledWith('cb-existing');
   });
+
+  it('should pass original platform conversation ID through to workflowRunCommand', async () => {
+    const workflowDb = await import('@archon/core/db/workflows');
+    const codebaseDb = await import('@archon/core/db/codebases');
+    const conversationsDb = await import('@archon/core/db/conversations');
+    const workflowDiscovery = await import('@archon/workflows/workflow-discovery');
+    const core = await import('@archon/core');
+
+    (workflowDb.getWorkflowRun as ReturnType<typeof mock>).mockResolvedValueOnce({
+      id: 'run-approve-conv',
+      workflow_name: 'implement',
+      status: 'paused',
+      user_message: 'add auth',
+      working_path: '/tmp/test-worktree',
+      codebase_id: 'cb-existing',
+      conversation_id: 'db-uuid-original',
+      metadata: { approval: { nodeId: 'review-node', message: 'Approve?' } },
+    });
+
+    // Return a conversation with the original platform ID
+    (conversationsDb.getConversationById as ReturnType<typeof mock>).mockResolvedValueOnce({
+      id: 'db-uuid-original',
+      platform_type: 'cli',
+      platform_conversation_id: 'cli-original-123',
+    });
+
+    (
+      workflowDiscovery.discoverWorkflowsWithConfig as ReturnType<typeof mock>
+    ).mockResolvedValueOnce({
+      workflows: [makeTestWorkflowWithSource({ name: 'implement' })],
+      errors: [],
+    });
+
+    (codebaseDb.getCodebase as ReturnType<typeof mock>).mockResolvedValueOnce({
+      id: 'cb-existing',
+      name: 'owner/repo',
+      default_cwd: '/path/to/main-checkout',
+    });
+
+    // Clear call history before our test so we can assert precisely
+    (conversationsDb.getOrCreateConversation as ReturnType<typeof mock>).mockClear();
+
+    try {
+      await workflowApproveCommand('run-approve-conv');
+    } catch {
+      // downstream failure is acceptable — we only need to reach getOrCreateConversation
+    }
+
+    // Verify the original platform conversation ID was passed through
+    expect(conversationsDb.getConversationById).toHaveBeenCalledWith('db-uuid-original');
+    expect(conversationsDb.getOrCreateConversation).toHaveBeenCalledWith('cli', 'cli-original-123');
+  });
 });
 
 describe('workflowAbandonCommand', () => {
@@ -1566,6 +1619,61 @@ describe('workflowRejectCommand', () => {
     expect(consoleSpy).toHaveBeenCalledWith(expect.stringContaining('Rejected workflow'));
   });
 
+  it('should pass original platform conversation ID through on reject-resume', async () => {
+    const workflowDb = await import('@archon/core/db/workflows');
+    const conversationsDb = await import('@archon/core/db/conversations');
+    const workflowDiscovery = await import('@archon/workflows/workflow-discovery');
+
+    const runData = {
+      id: 'run-reject-conv',
+      workflow_name: 'my-wf',
+      status: 'paused',
+      user_message: 'build it',
+      working_path: '/repo',
+      codebase_id: null,
+      conversation_id: 'db-uuid-reject',
+      metadata: {
+        approval: {
+          type: 'approval',
+          nodeId: 'gate',
+          message: 'Approve?',
+          onRejectPrompt: 'Fix: $REJECTION_REASON',
+          onRejectMaxAttempts: 3,
+        },
+        rejection_count: 0,
+      },
+    };
+    // rejectWorkflow reads the run twice internally (getRunOrThrow + updateWorkflowRun check)
+    (workflowDb.getWorkflowRun as ReturnType<typeof mock>).mockResolvedValueOnce(runData);
+
+    // Return a conversation with the original platform ID
+    (conversationsDb.getConversationById as ReturnType<typeof mock>).mockResolvedValueOnce({
+      id: 'db-uuid-reject',
+      platform_type: 'cli',
+      platform_conversation_id: 'cli-reject-456',
+    });
+
+    (
+      workflowDiscovery.discoverWorkflowsWithConfig as ReturnType<typeof mock>
+    ).mockResolvedValueOnce({
+      workflows: [makeTestWorkflowWithSource({ name: 'my-wf' })],
+      errors: [],
+    });
+
+    // Clear call history before our test so we can assert precisely
+    (conversationsDb.getOrCreateConversation as ReturnType<typeof mock>).mockClear();
+
+    try {
+      await workflowRejectCommand('run-reject-conv', 'needs work');
+    } catch {
+      // downstream workflowRunCommand failure is acceptable — we only need to reach getOrCreateConversation
+    }
+
+    // Verify the original platform conversation ID was passed through
+    expect(conversationsDb.getConversationById).toHaveBeenCalledWith('db-uuid-reject');
+    expect(conversationsDb.getOrCreateConversation).toHaveBeenCalledWith('cli', 'cli-reject-456');
+  });
+
   it('cancels when max attempts reached', async () => {
     const workflowDb = await import('@archon/core/db/workflows');
     const core = await import('@archon/core');
diff --git a/packages/cli/src/commands/workflow.ts b/packages/cli/src/commands/workflow.ts
index 99edcb0bfe..89dd5911e4 100644
--- a/packages/cli/src/commands/workflow.ts
+++ b/packages/cli/src/commands/workflow.ts
@@ -66,6 +66,8 @@ export interface WorkflowRunOptions {
   allowEnvKeys?: boolean;
   quiet?: boolean;
   verbose?: boolean;
+  /** Platform conversation ID (e.g. `cli-{ts}-{rand}`), NOT a DB UUID. */
+  conversationId?: string;
 }
 
 /**
@@ -269,7 +271,7 @@ export async function workflowRunCommand(
   const adapter = new CLIAdapter();
 
   // Generate conversation ID
-  const conversationId = generateConversationId();
+  const conversationId = options.conversationId ?? generateConversationId();
 
   // Get or create conversation in database
   let conversation;
@@ -861,10 +863,30 @@ export async function workflowApproveCommand(runId: string, comment?: string): P
   console.log('');
   console.log('Resuming workflow...');
 
+  // Look up the original platform conversation ID to keep all messages in one thread
+  let platformConversationId: string | undefined;
+  try {
+    const originalConversation = await conversationDb.getConversationById(result.conversationId);
+    platformConversationId = originalConversation?.platform_conversation_id ?? undefined;
+    if (!originalConversation) {
+      getLog().info(
+        { runId, conversationId: result.conversationId },
+        'cli.workflow_approve_conversation_not_found'
+      );
+    }
+  } catch (error) {
+    const err = error as Error;
+    getLog().warn(
+      { err, runId, conversationId: result.conversationId },
+      'cli.workflow_approve_conversation_lookup_failed'
+    );
+  }
+
   try {
     await workflowRunCommand(result.workingPath, result.workflowName, result.userMessage ?? '', {
       resume: true,
       codebaseId: result.codebaseId ?? undefined,
+      conversationId: platformConversationId,
     });
   } catch (error) {
     const err = error as Error;
@@ -900,10 +922,31 @@ export async function workflowRejectCommand(runId: string, reason?: string): Pro
   }
   console.log(`Rejected workflow: ${result.workflowName}`);
   console.log('Resuming with on_reject prompt...');
+
+  // Look up the original platform conversation ID to keep all messages in one thread
+  let platformConversationId: string | undefined;
+  try {
+    const originalConversation = await conversationDb.getConversationById(result.conversationId);
+    platformConversationId = originalConversation?.platform_conversation_id ?? undefined;
+    if (!originalConversation) {
+      getLog().info(
+        { runId, conversationId: result.conversationId },
+        'cli.workflow_reject_conversation_not_found'
+      );
+    }
+  } catch (error) {
+    const err = error as Error;
+    getLog().warn(
+      { err, runId, conversationId: result.conversationId },
+      'cli.workflow_reject_conversation_lookup_failed'
+    );
+  }
+
   try {
     await workflowRunCommand(result.workingPath, result.workflowName, result.userMessage ?? '', {
       resume: true,
       codebaseId: result.codebaseId ?? undefined,
+      conversationId: platformConversationId,
     });
   } catch (error) {
     const err = error as Error;
diff --git a/packages/core/src/operations/workflow-operations.ts b/packages/core/src/operations/workflow-operations.ts
index 8639ff1654..2b9092e649 100644
--- a/packages/core/src/operations/workflow-operations.ts
+++ b/packages/core/src/operations/workflow-operations.ts
@@ -34,6 +34,8 @@ export interface ApprovalOperationResult {
   workingPath: string | null;
   userMessage: string | null;
   codebaseId: string | null;
+  /** Internal DB UUID — resolve via getConversationById() to get platform_conversation_id. */
+  conversationId: string;
   type: 'interactive_loop' | 'approval_gate';
 }
 
@@ -42,6 +44,8 @@ export interface RejectionOperationResult {
   workingPath: string | null;
   userMessage: string | null;
   codebaseId: string | null;
+  /** Internal DB UUID — resolve via getConversationById() to get platform_conversation_id. */
+  conversationId: string;
   /** true = run cancelled; false = transitioning to failed for retry (has onRejectPrompt) */
   cancelled: boolean;
   /** true when cancelled specifically because max rejection attempts were reached */
@@ -168,6 +172,7 @@ export async function approveWorkflow(
         workingPath: run.working_path,
         userMessage: run.user_message,
         codebaseId: run.codebase_id,
+        conversationId: run.conversation_id,
         type: 'interactive_loop',
       };
     }
@@ -204,6 +209,7 @@ export async function approveWorkflow(
     workingPath: run.working_path,
     userMessage: run.user_message,
     codebaseId: run.codebase_id,
+    conversationId: run.conversation_id,
     type: 'approval_gate',
   };
 }
@@ -248,6 +254,7 @@ export async function rejectWorkflow(
           workingPath: run.working_path,
           userMessage: run.user_message,
           codebaseId: run.codebase_id,
+          conversationId: run.conversation_id,
           cancelled: true,
           maxAttemptsReached: true,
         };
@@ -261,6 +268,7 @@ export async function rejectWorkflow(
         workingPath: run.working_path,
         userMessage: run.user_message,
         codebaseId: run.codebase_id,
+        conversationId: run.conversation_id,
         cancelled: false,
         maxAttemptsReached: false,
       };
@@ -280,6 +288,7 @@ export async function rejectWorkflow(
     workingPath: run.working_path,
     userMessage: run.user_message,
     codebaseId: run.codebase_id,
+    conversationId: run.conversation_id,
     cancelled: true,
     maxAttemptsReached: false,
   };