story-kit: mergemaster conflict resolution, vite proxy fix, bug 279

- Upgrade mergemaster prompt to resolve complex conflicts itself
  instead of just reporting failure. Includes instructions to check
  git history and story files for context before resolving.
- Add proxy error handler to vite config to prevent crashes on
  backend ECONNREFUSED.
- Fix bug 279: auto-assign now checks that preferred agent's stage
  matches the pipeline stage. Coders won't be assigned to QA/merge.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

.story_kit/project.toml

@@ -220,7 +220,7 @@ role = "Merges completed coder work into master, runs quality gates, archives st
 model = "opus"
 max_turns = 30
 max_budget_usd = 5.00
-prompt = """You are the mergemaster agent for story {{story_id}}. Your job is to merge the completed coder work into master using the merge_agent_work MCP tool.
+prompt = """You are the mergemaster agent for story {{story_id}}. Your job is to merge the completed coder work into master.
 
 Read CLAUDE.md first, then .story_kit/README.md to understand the dev process.
 
@@ -229,20 +229,43 @@ Read CLAUDE.md first, then .story_kit/README.md to understand the dev process.
 2. Review the result: check success, had_conflicts, conflicts_resolved, gates_passed, and gate_output
 3. If merge succeeded and gates passed: report success to the human
 4. If conflicts were auto-resolved (conflicts_resolved=true) and gates passed: report success, noting which conflicts were resolved
-5. If conflicts could not be auto-resolved: call report_merge_failure(story_id='{{story_id}}', reason='<conflict details>') and report to the human. Master is untouched.
-6. If merge failed for any other reason: call report_merge_failure(story_id='{{story_id}}', reason='<details>') and report to the human.
-7. If gates failed after merge: attempt to fix minor issues (see below), then re-trigger merge_agent_work. After 2 fix attempts, call report_merge_failure and stop.
+5. If conflicts could not be auto-resolved: **resolve them yourself** in the merge worktree (see below)
+6. If merge failed for any other reason: call report_merge_failure(story_id='{{story_id}}', reason='<details>') and report to the human
+7. If gates failed after merge: attempt to fix the issues yourself in the merge worktree, then re-trigger merge_agent_work. After 3 fix attempts, call report_merge_failure and stop.
 
-## How Conflict Resolution Works
-The merge pipeline uses a temporary merge-queue branch and worktree to isolate merges from master. Simple additive conflicts (both branches adding code at the same location) are resolved automatically by keeping both additions. Complex conflicts (modifying the same lines differently) are reported without touching master.
+## Resolving Complex Conflicts Yourself
 
-## Fixing Minor Gate Failures
-If quality gates fail (cargo clippy, cargo test, npm run build, npm test), attempt to fix minor issues yourself before reporting to the human.
+When the auto-resolver fails, you have access to the merge worktree at `.story_kit/merge_workspace/`. Go in there and resolve the conflicts manually:
 
-**Fix yourself (up to 2 attempts total):**
+1. Run `git diff --name-only --diff-filter=U` in the merge worktree to list conflicted files
+2. **Build context before touching code.** Run `git log --oneline master...HEAD` on the feature branch to see its commits. Then run `git log --oneline --since="$(git log -1 --format=%ci <feature-branch-base-commit>)" master` to see what landed on master since the branch was created. Read the story files in `.story_kit/work/` for any recently merged stories that touch the same files — this tells you WHY master changed and what must be preserved.
+3. Read each conflicted file and understand both sides of the conflict
+4. **Understand intent, not just syntax.** The feature branch may be behind master — master's version of shared infrastructure is almost always correct. The feature branch's contribution is the NEW functionality it adds. Your job is to integrate the new into master's structure, not pick one side.
+5. Resolve by integrating the feature's new functionality into master's code structure
+5. Stage resolved files with `git add`
+6. Run `cargo check` (and `npm run build` if frontend changed) to verify compilation
+7. If it compiles, commit and re-trigger merge_agent_work
+
+### Common conflict patterns in this project:
+
+**Story file rename/rename conflicts:** Both branches moved the story .md file to different pipeline directories. Resolution: `git rm` both sides — story files in `work/2_current/`, `work/3_qa/`, `work/4_merge/` are gitignored and don't need to be committed.
+
+**bot.rs tokio::select! conflicts:** Master has a `tokio::select!` loop in `handle_message()` that handles permission forwarding (story 275). Feature branches created before story 275 have a simpler direct `provider.chat_stream().await` call. Resolution: KEEP master's tokio::select! loop. Integrate only the feature's new logic (e.g. typing indicators, new callbacks) into the existing loop structure. Do NOT replace the loop with the old direct call.
+
+**Duplicate functions/imports:** The auto-resolver keeps both sides, producing duplicates. Resolution: keep one copy (prefer master's version), delete the duplicate.
+
+**Formatting-only conflicts:** Both sides reformatted the same code differently. Resolution: pick either side (prefer master).
+
+## Fixing Gate Failures
+
+If quality gates fail (cargo clippy, cargo test, npm run build, npm test), attempt to fix issues yourself in the merge worktree.
+
+**Fix yourself (up to 3 attempts total):**
 - Syntax errors (missing semicolons, brackets, commas)
+- Duplicate definitions from merge artifacts
 - Simple type annotation errors
 - Unused import warnings flagged by clippy
+- Mismatched braces from bad conflict resolution
 - Trivial formatting issues that block compilation or linting
 
 **Report to human without attempting a fix:**
@@ -250,17 +273,14 @@ If quality gates fail (cargo clippy, cargo test, npm run build, npm test), attem
 - Missing function implementations
 - Architectural changes required
 - Non-trivial refactoring needed
-- Anything requiring understanding of broader system context
 
-**Max retry limit:** If gates still fail after 2 fix attempts, call report_merge_failure to record the failure, then stop immediately and report the full gate output to the human. Do not retry further.
+**Max retry limit:** If gates still fail after 3 fix attempts, call report_merge_failure to record the failure, then stop immediately and report the full gate output to the human.
 
 ## CRITICAL Rules
 - NEVER manually move story files between pipeline stages (e.g. from 4_merge/ to 5_done/)
 - NEVER call accept_story — only merge_agent_work can move stories to done after a successful merge
-- When merge fails, ALWAYS call report_merge_failure to record the failure — do NOT improvise with file moves
-- Only use MCP tools (merge_agent_work, report_merge_failure) to drive the merge process
-- Only attempt fixes that are clearly minor and low-risk
+- When merge fails after exhausting your fix attempts, ALWAYS call report_merge_failure
 - Report conflict resolution outcomes clearly
 - Report gate failures with full output so the human can act if needed
 - The server automatically runs acceptance gates when your process exits"""
-system_prompt = "You are the mergemaster agent. Your primary responsibility is to trigger the merge_agent_work MCP tool and report the results. CRITICAL: Never manually move story files or call accept_story. When merge fails, call report_merge_failure to record the failure. For minor gate failures (syntax errors, unused imports, missing semicolons), attempt to fix them yourself — but stop after 2 attempts, call report_merge_failure, and report to the human. For complex failures or unresolvable conflicts, call report_merge_failure and report clearly so the human can act. The merge pipeline automatically resolves simple additive conflicts."
+system_prompt = "You are the mergemaster agent. Your primary job is to merge feature branches to master. First try the merge_agent_work MCP tool. If the auto-resolver fails on complex conflicts, resolve them yourself in the merge worktree — you are an opus-class agent capable of understanding both sides of a conflict and producing correct merged code. Common patterns: keep master's tokio::select! permission loop in bot.rs, discard story file rename conflicts (gitignored), remove duplicate definitions. After resolving, verify compilation before re-triggering merge. CRITICAL: Never manually move story files or call accept_story. After 3 failed fix attempts, call report_merge_failure and stop."

frontend/vite.config.ts

@@ -16,6 +16,12 @@ export default defineConfig(() => {
         "/api": {
           target: `http://127.0.0.1:${String(backendPort)}`,
           timeout: 120000,
+          configure: (proxy) => {
+            proxy.on("error", (_err) => {
+              // Swallow proxy errors (e.g. ECONNREFUSED during backend restart)
+              // so the vite dev server doesn't crash.
+            });
+          },
         },
       },
       watch: {

server/src/agents/pool.rs

@@ -1468,10 +1468,12 @@ impl AgentPool {
                 let preferred_agent =
                     read_story_front_matter_agent(project_root, stage_dir, story_id);
 
-                // Outcome: (already_assigned, chosen_agent, preferred_busy)
+                // Outcome: (already_assigned, chosen_agent, preferred_busy, stage_mismatch)
                 // preferred_busy=true means the story has a specific agent requested but it is
                 // currently occupied — the story should wait rather than fall back.
-                let (already_assigned, free_agent, preferred_busy) = {
+                // stage_mismatch=true means the preferred agent's stage doesn't match the
+                // pipeline stage, so we fell back to a generic stage agent.
+                let (already_assigned, free_agent, preferred_busy, stage_mismatch) = {
                     let agents = match self.agents.lock() {
                         Ok(a) => a,
                         Err(e) => {
@@ -1481,18 +1483,29 @@ impl AgentPool {
                     };
                     let assigned = is_story_assigned_for_stage(&config, &agents, story_id, stage);
                     if assigned {
-                        (true, None, false)
+                        (true, None, false, false)
                     } else if let Some(ref pref) = preferred_agent {
                         // Story has a front-matter agent preference.
-                        if is_agent_free(&agents, pref) {
-                            (false, Some(pref.clone()), false)
+                        // Verify the preferred agent's stage matches the current
+                        // pipeline stage — a coder shouldn't be assigned to QA.
+                        let pref_stage_matches = config
+                            .find_agent(pref)
+                            .map(|cfg| agent_config_stage(cfg) == *stage)
+                            .unwrap_or(false);
+                        if !pref_stage_matches {
+                            // Stage mismatch — fall back to any free agent for this stage.
+                            let free = find_free_agent_for_stage(&config, &agents, stage)
+                                .map(|s| s.to_string());
+                            (false, free, false, true)
+                        } else if is_agent_free(&agents, pref) {
+                            (false, Some(pref.clone()), false, false)
                         } else {
-                            (false, None, true)
+                            (false, None, true, false)
                         }
                     } else {
                         let free = find_free_agent_for_stage(&config, &agents, stage)
                             .map(|s| s.to_string());
-                        (false, free, false)
+                        (false, free, false, false)
                     }
                 };
 
@@ -1511,6 +1524,13 @@ impl AgentPool {
                     continue;
                 }
 
+                if stage_mismatch {
+                    slog!(
+                        "[auto-assign] Preferred agent '{}' stage mismatch for '{story_id}' in {stage_dir}/; falling back to stage-appropriate agent.",
+                        preferred_agent.as_deref().unwrap_or("?")
+                    );
+                }
+
                 match free_agent {
                     Some(agent_name) => {
                         slog!(
@@ -4748,4 +4768,130 @@ stage = "coder"
             "No agents should be assigned to a spike with review_hold"
         );
     }
+
+    // ── Story 279: auto-assign respects agent stage from front matter ──────────
+
+    /// When a story in 3_qa/ has `agent: coder-1` in its front matter but
+    /// coder-1 is a coder-stage agent, auto-assign must NOT assign coder-1.
+    /// Instead it should fall back to a free QA-stage agent.
+    #[tokio::test]
+    async fn auto_assign_ignores_coder_preference_when_story_is_in_qa_stage() {
+        let tmp = tempfile::tempdir().unwrap();
+        let sk = tmp.path().join(".story_kit");
+        let qa_dir = sk.join("work/3_qa");
+        std::fs::create_dir_all(&qa_dir).unwrap();
+        std::fs::write(
+            sk.join("project.toml"),
+            "[[agent]]\nname = \"coder-1\"\nstage = \"coder\"\n\n\
+             [[agent]]\nname = \"qa-1\"\nstage = \"qa\"\n",
+        )
+        .unwrap();
+        // Story in 3_qa/ with a preferred coder-stage agent.
+        std::fs::write(
+            qa_dir.join("story-qa1.md"),
+            "---\nname: QA Story\nagent: coder-1\n---\n",
+        )
+        .unwrap();
+
+        let pool = AgentPool::new_test(3001);
+
+        pool.auto_assign_available_work(tmp.path()).await;
+
+        let agents = pool.agents.lock().unwrap();
+        // coder-1 must NOT have been assigned (wrong stage for 3_qa/).
+        let coder_assigned = agents
+            .values()
+            .any(|a| a.agent_name == "coder-1" && matches!(a.status, AgentStatus::Pending | AgentStatus::Running));
+        assert!(
+            !coder_assigned,
+            "coder-1 should not be assigned to a QA-stage story"
+        );
+        // qa-1 should have been assigned instead.
+        let qa_assigned = agents
+            .values()
+            .any(|a| a.agent_name == "qa-1" && matches!(a.status, AgentStatus::Pending | AgentStatus::Running));
+        assert!(
+            qa_assigned,
+            "qa-1 should be assigned as fallback for the QA-stage story"
+        );
+    }
+
+    /// When a story in 2_current/ has `agent: coder-1` in its front matter and
+    /// coder-1 is a coder-stage agent, auto-assign must respect the preference
+    /// and assign coder-1 (not fall back to some other coder).
+    #[tokio::test]
+    async fn auto_assign_respects_coder_preference_when_story_is_in_current_stage() {
+        let tmp = tempfile::tempdir().unwrap();
+        let sk = tmp.path().join(".story_kit");
+        let current_dir = sk.join("work/2_current");
+        std::fs::create_dir_all(&current_dir).unwrap();
+        std::fs::write(
+            sk.join("project.toml"),
+            "[[agent]]\nname = \"coder-1\"\nstage = \"coder\"\n\n\
+             [[agent]]\nname = \"coder-2\"\nstage = \"coder\"\n",
+        )
+        .unwrap();
+        // Story in 2_current/ with a preferred coder-1 agent.
+        std::fs::write(
+            current_dir.join("story-pref.md"),
+            "---\nname: Coder Story\nagent: coder-1\n---\n",
+        )
+        .unwrap();
+
+        let pool = AgentPool::new_test(3001);
+
+        pool.auto_assign_available_work(tmp.path()).await;
+
+        let agents = pool.agents.lock().unwrap();
+        // coder-1 should have been picked (it matches the stage and is preferred).
+        let coder1_assigned = agents
+            .values()
+            .any(|a| a.agent_name == "coder-1" && matches!(a.status, AgentStatus::Pending | AgentStatus::Running));
+        assert!(
+            coder1_assigned,
+            "coder-1 should be assigned when it matches the stage and is preferred"
+        );
+        // coder-2 must NOT be assigned (not preferred).
+        let coder2_assigned = agents
+            .values()
+            .any(|a| a.agent_name == "coder-2" && matches!(a.status, AgentStatus::Pending | AgentStatus::Running));
+        assert!(
+            !coder2_assigned,
+            "coder-2 should not be assigned when coder-1 is explicitly preferred"
+        );
+    }
+
+    /// When the preferred agent's stage mismatches and no other agent of the
+    /// correct stage is available, auto-assign must not start any agent for that
+    /// story (no panic, no error).
+    #[tokio::test]
+    async fn auto_assign_stage_mismatch_with_no_fallback_starts_no_agent() {
+        let tmp = tempfile::tempdir().unwrap();
+        let sk = tmp.path().join(".story_kit");
+        let qa_dir = sk.join("work/3_qa");
+        std::fs::create_dir_all(&qa_dir).unwrap();
+        // Only a coder agent is configured — no QA agent exists.
+        std::fs::write(
+            sk.join("project.toml"),
+            "[[agent]]\nname = \"coder-1\"\nstage = \"coder\"\n",
+        )
+        .unwrap();
+        // Story in 3_qa/ requests coder-1 (wrong stage) and no QA agent exists.
+        std::fs::write(
+            qa_dir.join("story-noqa.md"),
+            "---\nname: QA Story No Agent\nagent: coder-1\n---\n",
+        )
+        .unwrap();
+
+        let pool = AgentPool::new_test(3001);
+
+        // Must not panic.
+        pool.auto_assign_available_work(tmp.path()).await;
+
+        let agents = pool.agents.lock().unwrap();
+        assert!(
+            agents.is_empty(),
+            "No agent should be started when no stage-appropriate agent is available"
+        );
+    }
 }