fix(lint): measure the Go lint debt honestly, and clear the production half

.github/workflows/ci.yml

@@ -124,8 +124,13 @@ jobs:
         with:
           version: latest
           working-directory: go
-          # AX: Skip test files (covered by test-go job), timeout for large codebase
-          args: --timeout=5m --tests=false
+          # Tests are linted, not skipped. --tests=false was hiding two things
+          # at once: it reported 21 production symbols as unused because the
+          # only callers are in _test.go files — every one of them a
+          # test-injection seam like syncPull, mcpInitialize or newCoreAgent —
+          # while never reporting the dead scaffolding actually in the test
+          # files. Linting tests swaps 21 false positives for 28 real ones.
+          args: --timeout=5m
 
   # Guards the one rule that has silently cost delivered features: the sibling
   # dappco.re/* modules are dependencies, not workspace files.

go/.golangci.yml

@@ -0,0 +1,35 @@
+# golangci-lint configuration.
+#
+# Policy lived in a CI argument (--tests=false) until now, which is why it was
+# both wrong and invisible. Two problems it caused:
+#
+#   * Skipping test files made `unused` report 21 production symbols as dead
+#     because their only callers are in _test.go — every one a deliberate
+#     test-injection seam (syncPull, mcpInitialize, mcpCall, readSSEData,
+#     newCoreAgent, pokeCh and friends). The linter cannot see a caller it has
+#     been told not to read.
+#   * It simultaneously hid the dead scaffolding actually inside the test
+#     files, which is where all 28 genuine `unused` findings turn out to live.
+#
+# So tests are linted. errcheck is excluded for them instead, which is the
+# narrower and honest cut: an unchecked Close() in test setup is noise, while
+# an unchecked write-close in production can silently lose data — as two did in
+# pkg/chathistory before this sweep.
+version: "2"
+
+linters:
+  exclusions:
+    rules:
+      # Unchecked errors in tests are conventional and not worth the churn;
+      # everything else still applies to them, which is the point.
+      - path: '_test\.go'
+        linters:
+          - errcheck
+
+issues:
+  # Report everything. The defaults are max-issues-per-linter: 50 and
+  # max-same-issues: 3, which silently truncate — the debt in this repo read as
+  # "90 issues" for exactly as long as nobody passed these, and is really 839.
+  # A capped number is not a measurement.
+  max-issues-per-linter: 0
+  max-same-issues: 0

go/cmd/core-agent/commands_chat.go

@@ -53,7 +53,13 @@ func (commands applicationCommandSet) chat(opts core.Options) core.Result {
 		applicationPrint("chat: open archive: %v", err)
 		return core.Result{}
 	}
-	defer hist.Close()
+	defer func() {
+		// Reported, not dropped: this handle is written to, so a failed
+		// close can mean the last turns never reached the archive.
+		if err := hist.Close(); err != nil {
+			core.Warn("chat: failed to close history archive", "reason", err)
+		}
+	}()
 
 	svc := lemma.New(lemma.Config{
 		BaseURL: baseURL,
@@ -78,7 +84,9 @@ func (commands applicationCommandSet) chat(opts core.Options) core.Result {
 	scanner := bufio.NewScanner(core.Stdin())
 	scanner.Buffer(make([]byte, 64*1024), 1024*1024) // allow long prompts
 	for {
-		core.WriteString(stdout, "you: ")
+		// Prompt only: a failed terminal write is cosmetic and there is
+		// nowhere to return an error to inside the loop.
+		_ = core.WriteString(stdout, "you: ")
 		if !scanner.Scan() {
 			break
 		}

go/cmd/core-agent/lemma_mcp.go

@@ -113,7 +113,13 @@ func (s *lemmaSubsystem) handleSend(ctx context.Context, input LemmaSendInput) (
 	if err != nil {
 		return nil, LemmaSendOutput{}, err
 	}
-	defer hist.Close()
+	defer func() {
+		// Reported, not dropped: this handle is written to, so a failed
+		// close can mean the last turns never reached the archive.
+		if err := hist.Close(); err != nil {
+			core.Warn("chat: failed to close history archive", "reason", err)
+		}
+	}()
 
 	cfg := s.cfg
 	cfg.History = hist

go/cmd/core-agent/main.go

@@ -125,7 +125,14 @@ var runApp = func(coreApp *core.Core, cliArgs []string) error {
 		return core.E("main.runApp", "core is required", nil)
 	}
 
-	defer coreApp.ServiceShutdown(context.Background())
+	defer func() {
+		// Reported on the way out: a shutdown that fails can leave a
+		// socket or database handle behind, and this is the last chance
+		// anything gets to say so.
+		if r := coreApp.ServiceShutdown(context.Background()); !r.OK {
+			core.Warn("core-agent: service shutdown failed", "reason", r.Value)
+		}
+	}()
 
 	result := coreApp.ServiceStartup(coreApp.Context(), nil)
 	if !result.OK {

go/pkg/agentic/auto_pr.go

@@ -93,7 +93,9 @@ func (s *PrepSubsystem) autoCreatePR(workspaceDir string) {
 			return
 		}
 		workspaceStatusUpdate.PRURL = pullRequestURL
-		writeStatusResult(workspaceDir, workspaceStatusUpdate)
+		if r := writeStatusResult(workspaceDir, workspaceStatusUpdate); !r.OK {
+			core.Warn("agentic: failed to record PR URL on status", "workspace", workspaceDir, "reason", r.Value)
+		}
 	}
 }
 

go/pkg/agentic/dispatch.go

@@ -654,8 +654,9 @@
 		}); !result.OK {
 			core.Warn("agentic.broadcastComplete: notification failed", "reason", result.Error())
 		}
-		// Push to MCP channel so Claude Code receives the notification
-		s.Core().ACTION(coremcp.ChannelPush{
+		// Push to MCP channel so Claude Code receives the notification.
+		// Best-effort: a listener that is gone must not fail the dispatch.
+		_ = s.Core().ACTION(coremcp.ChannelPush{
 			Channel: coremcp.ChannelAgentComplete,
 			Data: map[string]any{
 				"agent": agent, "repo": repo,
@@ -667,7 +668,7 @@
 
 func (s *PrepSubsystem) onAgentComplete(agent, workspaceDir, outputFile string, exitCode int, processStatus, output string) {
 	if output != "" {
 		fs.Write(outputFile, output)
 	}
 
 	repoDir := WorkspaceRepoDir(workspaceDir)
@@ -686,7 +687,9 @@
 		workspaceStatus.Status = finalStatus
 		workspaceStatus.PID = 0
 		workspaceStatus.Question = question
-		writeStatusResult(workspaceDir, workspaceStatus)
+		if r := writeStatusResult(workspaceDir, workspaceStatus); !r.OK {
+			core.Warn("agentic: failed to write blocked status", "workspace", workspaceDir, "reason", r.Value)
+		}
 		s.TrackWorkspace(WorkspaceName(workspaceDir), workspaceStatus)
 
 		s.trackFailureRate(agent, finalStatus, workspaceStatus.StartedAt)
@@ -891,7 +894,9 @@
 				StartedAt: time.Now(),
 				Runs:      0,
 			}
-			writeStatusResult(workspaceDir, workspaceStatus)
+			if r := writeStatusResult(workspaceDir, workspaceStatus); !r.OK {
+				core.Warn("agentic: failed to write requeued status", "workspace", workspaceDir, "reason", r.Value)
+			}
 			if runnerResult := s.Core().Service("runner"); runnerResult.OK {
 				if runnerSvc, ok := runnerResult.Value.(workspaceTracker); ok {
 					runnerSvc.TrackWorkspace(WorkspaceName(workspaceDir), workspaceStatus)
@@ -925,7 +930,7 @@
 		Runs:      1,
 	}
 	preserveStatusNote(workspaceDir, workspaceStatus) // keep VZ→OCI downgrade note (SP2.4)
 	writeStatusResult(workspaceDir, workspaceStatus)
 	if s.ServiceRuntime != nil {
 		if runnerResult := s.Core().Service("runner"); runnerResult.OK {
 			if runnerSvc, ok := runnerResult.Value.(workspaceTracker); ok {

go/pkg/brain/messaging.go

@@ -121,7 +121,9 @@ func (s *DirectSubsystem) notifySelf(ctx context.Context, input SendInput) {
 		return
 	}
 	for session := range mcpSvc.Sessions() {
-		coremcp.NotifySession(ctx, session, "notifications/claude/channel", map[string]any{
+		// Best-effort per session: one unreachable listener must not stop
+		// the loop notifying the rest.
+		_ = coremcp.NotifySession(ctx, session, "notifications/claude/channel", map[string]any{
 			"content": input.Content,
 			"meta": map[string]string{
 				"from":    agentic.AgentName(),

go/pkg/chathistory/chathistory.go

@@ -288,7 +288,7 @@ func (h *History) LoadTurns(conversationID string) ([]Turn, error) {
 	if err != nil {
 		return nil, core.E("chathistory.LoadTurns", "query", err)
 	}
-	defer rows.Close()
+	defer func() { _ = rows.Close() }() // read cursor
 	var out []Turn
 	for rows.Next() {
 		var t Turn

go/pkg/chathistory/export.go

@@ -35,7 +35,7 @@ func (h *History) CopyTo(dest string) error {
 		return core.E("chathistory.CopyTo", "open source", srcResult.Value.(error))
 	}
 	src := srcResult.Value.(*core.OSFile)
-	defer src.Close()
+	defer func() { _ = src.Close() }() // read handle: nothing to lose on close
 	if dir := core.PathDir(dest); dir != "" {
 		if r := core.MkdirAll(dir, 0o755); !r.OK {
 			return core.E("chathistory.CopyTo", "mkdir dest parent", r.Value.(error))
@@ -46,10 +46,17 @@ func (h *History) CopyTo(dest string) error {
 		return core.E("chathistory.CopyTo", "create dest", dstResult.Value.(error))
 	}
 	dst := dstResult.Value.(*core.OSFile)
-	defer dst.Close()
+	// Net for the error paths below; the success path closes explicitly.
+	defer func() { _ = dst.Close() }()
 	if _, err := io.Copy(dst, src); err != nil {
 		return core.E("chathistory.CopyTo", "copy bytes", err)
 	}
+	// Checked, not deferred-and-forgotten: a write handle that fails to close
+	// may never have flushed, and returning nil there reports a copy that did
+	// not fully happen.
+	if err := dst.Close(); err != nil {
+		return core.E("chathistory.CopyTo", "close dest", err)
+	}
 	return nil
 }
 
@@ -107,7 +114,8 @@ func (h *History) ExportJSONL(dest string) error {
 		return core.E("chathistory.ExportJSONL", "create dest", fResult.Value.(error))
 	}
 	f := fResult.Value.(*core.OSFile)
-	defer f.Close()
+	// Net for the error paths below; the success path closes explicitly.
+	defer func() { _ = f.Close() }()
 
 	convRows, err := h.db.Query(
 		`SELECT id, user_id, title, started_at, ended_at, model_id, base_model,
@@ -118,7 +126,7 @@ func (h *History) ExportJSONL(dest string) error {
 	if err != nil {
 		return core.E("chathistory.ExportJSONL", "query conversations", err)
 	}
-	defer convRows.Close()
+	defer func() { _ = convRows.Close() }() // read cursor
 
 	for convRows.Next() {
 		var c JSONLConversation
@@ -163,7 +171,7 @@ func (h *History) ExportJSONL(dest string) error {
 				&toolCalls, &toolResults, &t.CreatedAt,
 				&tokensIn, &tokensOut, &signal,
 			); err != nil {
-				turnRows.Close()
+				_ = turnRows.Close() // read cursor; the scan error is what matters
 				return core.E("chathistory.ExportJSONL", "scan turn", err)
 			}
 			if toolCalls.Valid {
@@ -181,7 +189,7 @@ func (h *History) ExportJSONL(dest string) error {
 			t.Signal = signal.String
 			c.Turns = append(c.Turns, t)
 		}
-		turnRows.Close()
+		_ = turnRows.Close() // read cursor
 
 		marshalled := core.JSONMarshal(c)
 		if !marshalled.OK {
@@ -195,5 +203,13 @@ func (h *History) ExportJSONL(dest string) error {
 			return core.E("chathistory.ExportJSONL", "write newline", err)
 		}
 	}
+
+	// Checked, not deferred-and-forgotten: every line above can be written
+	// successfully and still be lost if the final flush fails, which would
+	// report a complete export of a truncated file.
+	if err := f.Close(); err != nil {
+		return core.E("chathistory.ExportJSONL", "close dest", err)
+	}
+
 	return nil
 }

go/pkg/lemma/admin.go

@@ -308,7 +308,7 @@ func (a *Admin) doJSON(ctx context.Context, method, path string, body, out any)
 	if err != nil {
 		return core.E("lemma.Admin.doJSON", "transport", err)
 	}
-	defer resp.Body.Close()
+	defer func() { _ = resp.Body.Close() }() // read side of the response
 
 	respBody, _ := io.ReadAll(io.LimitReader(resp.Body, 1<<20))
 	if resp.StatusCode >= 400 {

go/pkg/lemma/lemma.go

@@ -287,7 +287,7 @@ func (s *Service) callChatCompletions(ctx context.Context, messages []chatMessag
 	if err != nil {
 		return "", 0, 0, err
 	}
-	defer resp.Body.Close()
+	defer func() { _ = resp.Body.Close() }() // read side of the response
 
 	rawBody, err := io.ReadAll(resp.Body)
 	if err != nil {

go/pkg/monitor/harvest.go

@@ -43,12 +43,15 @@ func (m *Subsystem) harvestCompleted() string {
 		if h.rejected != "" {
 			parts = append(parts, core.Sprintf("%s: REJECTED (%s)", h.repo, h.rejected))
 			if m.ServiceRuntime != nil {
-				m.Core().ACTION(messages.HarvestRejected{Repo: h.repo, Branch: h.branch, Reason: h.rejected})
+				// Best-effort notification: a failed dispatch must not
+				// abort the harvest that produced it.
+				_ = m.Core().ACTION(messages.HarvestRejected{Repo: h.repo, Branch: h.branch, Reason: h.rejected})
 			}
 		} else {
 			parts = append(parts, core.Sprintf("%s: ready-for-review %s (%d files)", h.repo, h.branch, h.files))
 			if m.ServiceRuntime != nil {
-				m.Core().ACTION(messages.HarvestComplete{Repo: h.repo, Branch: h.branch, Files: h.files})
+				// Best-effort notification, as above.
+				_ = m.Core().ACTION(messages.HarvestComplete{Repo: h.repo, Branch: h.branch, Files: h.files})
 			}
 		}
 	}

go/pkg/monitor/monitor.go

@@ -245,7 +245,9 @@ func (m *Subsystem) checkIdleAfterDelay() {
 
 	running, queued := m.countLiveWorkspaces()
 	if running == 0 && queued == 0 {
-		m.Core().ACTION(messages.QueueDrained{Completed: 0})
+		// Best-effort: a listener that has gone away must not stop the
+		// monitor noticing the queue drained.
+		_ = m.Core().ACTION(messages.QueueDrained{Completed: 0})
 	}
 }
 
@@ -333,7 +335,9 @@ func (m *Subsystem) check(ctx context.Context) {
 	m.notify(ctx, combinedMessage)
 
 	if m.svc != nil {
-		m.svc.Server().ResourceUpdated(ctx, &mcp.ResourceUpdatedNotificationParams{
+		// Best-effort: a subscriber that cannot be notified must not
+		// stop the status update itself.
+		_ = m.svc.Server().ResourceUpdated(ctx, &mcp.ResourceUpdatedNotificationParams{
 			URI: "status://agents",
 		})
 	}
@@ -403,7 +407,9 @@ func (m *Subsystem) checkCompletions() string {
 
 	liveRunning, liveQueued := m.countLiveWorkspaces()
 	if m.ServiceRuntime != nil && liveRunning == 0 && liveQueued == 0 {
-		m.Core().ACTION(messages.QueueDrained{Completed: len(newlyCompleted)})
+		// Best-effort: a listener that has gone away must not stop the
+		// monitor noticing the queue drained.
+		_ = m.Core().ACTION(messages.QueueDrained{Completed: len(newlyCompleted)})
 	}
 
 	msg := core.Sprintf("%d agent(s) completed", len(newlyCompleted))
@@ -505,7 +511,9 @@ func (m *Subsystem) notify(ctx context.Context, message string) {
 	}
 
 	for session := range m.svc.Server().Sessions() {
-		session.Log(ctx, &mcp.LoggingMessageParams{
+		// Best-effort: one unreachable session must not stop the loop
+		// logging to the others.
+		_ = session.Log(ctx, &mcp.LoggingMessageParams{
 			Level:  "info",
 			Logger: "monitor",
 			Data:   message,

go/pkg/runner/runner.go

@@ -284,8 +284,10 @@ func (s *Service) TrackWorkspace(name string, status any) {
 	if workspaceStatus == nil {
 		return
 	}
-	s.workspaces.Set(name, workspaceStatus)
-	s.workspaces.Delete(core.Concat("pending/", workspaceStatus.Repo))
+	if r := s.workspaces.Set(name, workspaceStatus); !r.OK {
+		core.Warn("runner: failed to track workspace", "workspace", name, "reason", r.Value)
+	}
+	_ = s.workspaces.Delete(core.Concat("pending/", workspaceStatus.Repo)) // best-effort: the pending marker may already be gone
 }
 
 // s.Workspaces().Each(func(name string, workspaceStatus *WorkspaceStatus) { core.Println(name, workspaceStatus.Status) })