Merge pull request #304 from awf-project/feature/F078-fix-cli-provider-invocations-breaking-ch

feat(agents): fix CLI provider invocations and add streaming output support (F078)

Author: pocky

CHANGELOG.md

@@ -9,10 +9,12 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Breaking Changes
 
+- **F078**: CLI provider invocation flags updated to match current binary APIs — Claude and Gemini `output_format: json` now maps to `--output-format stream-json` (was `--output-format json`); Codex invocation changed from `codex --prompt "<prompt>" --quiet` to `codex exec --json "<prompt>"`; `quiet` option removed from Codex (silently ignored); Codex conversation resume changed from `codex resume <id> --prompt "<prompt>"` to `codex resume <id> --json "<prompt>"`; workflows using `output_format: json` require no YAML changes (mapping is automatic); workflows using `quiet: true` for Codex should remove the option (no-op)
 - **F077**: Option keys normalized to snake_case — `allowedTools` renamed to `allowed_tools`, `dangerouslySkipPermissions` renamed to `dangerously_skip_permissions` in workflow YAML; old camelCase keys are silently ignored (Go map miss); `dangerously_skip_permissions` fails closed (permissions not skipped), `allowed_tools` fails open (no tool restriction applied); update existing workflow files to use the new snake_case keys
 
 ### Added
 
+- **F078**: OpenCode `--model` flag support — `model` option in workflow YAML now passed as `--model <value>` to OpenCode CLI in both `Execute` and `ExecuteConversation`; OpenCode always passes `--format json` for structured output
 - **F077**: `dangerously_skip_permissions` support for Gemini (`--approval-mode=yolo`) and Codex (`--yolo`) providers — unified permission bypass key works across all three agent providers (Claude, Gemini, Codex)
 - **F076**: `awf upgrade` self-update command — checks latest release on GitHub, downloads platform-specific binary, verifies SHA256 checksum, and atomically replaces the current executable; `--check` reports available updates without installing; `--version v0.5.0` installs a specific version; `--force` upgrades even if already on latest or running a dev build; heuristic warning when binary appears managed by a package manager (homebrew, apt, snap, nix); cross-filesystem fallback (copy + chmod) when `os.Rename` fails; `GITHUB_TOKEN` env var supported for rate-limited environments
 
@@ -22,6 +24,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Removed
 
+- **F078**: Dead validation helpers `validatePrompt()`, `validateContext()`, `validateState()` removed from agent helpers — all were no-ops or unreachable after provider refactoring
 - **F077**: Dead helper functions `getWorkflowID()` and `getStepName()` removed from agent helpers — keys `workflowID`/`stepName` were never injected by any caller; `workflow` and `step` fields removed from Claude provider audit log (redundant with execution service context)
 
 ## [0.6.0] - 2026-04-05

CLAUDE.md

@@ -217,7 +217,6 @@ func TestWorkflowValidation(t *testing.T) {
 
 ## Architecture Rules
 
-- Apply bug fixes uniformly across all components implementing the same pattern; verify path resolution consistency across all executors when fixing one
 - Never modify production code in test-only fixes; bugs discovered during testing must be documented in Bug Escalation Protocol (.specify/implementation/ISSUE/bug/) before implementing fixes
 - Document discovered runtime bugs in .specify/implementation/ISSUE/bug/ directory before implementation; prevents scope creep and enables separate tracking from test fixes
 - Own timeout responsibility in application layer via context.WithTimeout; infrastructure adapters must respect context cancellation without enforcing additional timeouts
@@ -239,6 +238,7 @@ func TestWorkflowValidation(t *testing.T) {
 - Initialize ApproximationTokenizer immediately before NewConversationManager in interfaces layer; token counting must be ready before conversation context is established
 - Resolve user-provided interpolated variables before registry or dependency lookups to enable dynamic selection at runtime; apply identical resolution logic across all related code paths
 - Implement per-provider flag mapping without shared abstraction when CLI syntax diverges fundamentally; document divergence (Claude: --flag-name, Gemini: --flag-name=value, Codex: --flag-name) inline
+- Synchronize provider CLI flag changes across both implementation files and central options configuration (options.go); verify declarations and validation rules align
 
 ## Common Pitfalls
 
@@ -306,6 +306,3 @@ func TestWorkflowValidation(t *testing.T) {
 - Extract repeated test assertion patterns (>5 duplicates) into table-driven or closure-based helpers to eliminate code duplication
 - Extract HTTP server setup patterns from integration tests into helper functions; eliminate duplication across multiple test functions
 - When flipping integration test assertions for newly-enabled features, transition from 'not configured' errors to provider-level implementation errors; verify assertions change state, not disappear
-
-## Review Standards
-

internal/application/agent_step_test.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"errors"
 	"fmt"
+	"io"
 	"strings"
 	"testing"
 	"time"
@@ -186,7 +187,7 @@ func TestExecutionService_AgentStep_BasicExecution(t *testing.T) {
 
 	registry := mocks.NewMockAgentRegistry()
 	claude := mocks.NewMockAgentProvider("claude")
-	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any) (*workflow.AgentResult, error) {
+	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any, stdout, stderr io.Writer) (*workflow.AgentResult, error) {
 		if prompt == "Summarize this text" {
 			return &workflow.AgentResult{
 				Provider:    "claude",
@@ -265,7 +266,7 @@ func TestExecutionService_AgentStep_WithOnFailure(t *testing.T) {
 
 	registry := mocks.NewMockAgentRegistry()
 	claude := mocks.NewMockAgentProvider("claude")
-	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any) (*workflow.AgentResult, error) {
+	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any, stdout, stderr io.Writer) (*workflow.AgentResult, error) {
 		if prompt == "Summarize this text" {
 			return &workflow.AgentResult{
 				Provider:    "claude",
@@ -348,7 +349,7 @@ func TestExecutionService_AgentStep_InMixedWorkflow(t *testing.T) {
 
 	registry := mocks.NewMockAgentRegistry()
 	claude := mocks.NewMockAgentProvider("claude")
-	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any) (*workflow.AgentResult, error) {
+	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any, stdout, stderr io.Writer) (*workflow.AgentResult, error) {
 		if prompt == "Analyze the prepared data" {
 			return &workflow.AgentResult{
 				Provider:    "claude",
@@ -431,7 +432,7 @@ func TestExecutionService_AgentStep_StepTimeout(t *testing.T) {
 
 	registry := mocks.NewMockAgentRegistry()
 	claude := mocks.NewMockAgentProvider("claude")
-	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any) (*workflow.AgentResult, error) {
+	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any, stdout, stderr io.Writer) (*workflow.AgentResult, error) {
 		if prompt == "Summarize this text" {
 			return &workflow.AgentResult{
 				Provider:    "claude",
@@ -501,7 +502,7 @@ func TestExecutionService_AgentStep_AgentTimeout(t *testing.T) {
 
 	registry := mocks.NewMockAgentRegistry()
 	claude := mocks.NewMockAgentProvider("claude")
-	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any) (*workflow.AgentResult, error) {
+	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any, stdout, stderr io.Writer) (*workflow.AgentResult, error) {
 		if prompt == "Summarize this text" {
 			return &workflow.AgentResult{
 				Provider:    "claude",
@@ -574,7 +575,7 @@ func TestExecutionService_AgentStep_ContextCancellation(t *testing.T) {
 
 	registry := mocks.NewMockAgentRegistry()
 	claude := mocks.NewMockAgentProvider("claude")
-	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any) (*workflow.AgentResult, error) {
+	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any, stdout, stderr io.Writer) (*workflow.AgentResult, error) {
 		return nil, context.Canceled
 	})
 	_ = registry.Register(claude)
@@ -644,7 +645,7 @@ func TestExecutionService_AgentStep_InParallelBranches(t *testing.T) {
 	registry := mocks.NewMockAgentRegistry()
 
 	claude := mocks.NewMockAgentProvider("claude")
-	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any) (*workflow.AgentResult, error) {
+	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any, stdout, stderr io.Writer) (*workflow.AgentResult, error) {
 		if prompt == "Analyze sentiment" {
 			return &workflow.AgentResult{
 				Provider:    "claude",
@@ -667,7 +668,7 @@ func TestExecutionService_AgentStep_InParallelBranches(t *testing.T) {
 	_ = registry.Register(claude)
 
 	gemini := mocks.NewMockAgentProvider("gemini")
-	gemini.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any) (*workflow.AgentResult, error) {
+	gemini.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any, stdout, stderr io.Writer) (*workflow.AgentResult, error) {
 		if prompt == "Extract keywords" {
 			return &workflow.AgentResult{
 				Provider:    "gemini",
@@ -750,7 +751,7 @@ func TestExecutionService_AgentStep_PromptInterpolation(t *testing.T) {
 	registry := mocks.NewMockAgentRegistry()
 	claude := mocks.NewMockAgentProvider("claude")
 	// Note: mock resolver doesn't interpolate, so prompt stays as-is
-	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any) (*workflow.AgentResult, error) {
+	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any, stdout, stderr io.Writer) (*workflow.AgentResult, error) {
 		if prompt == "Explain {{inputs.topic}} in simple terms" {
 			return &workflow.AgentResult{
 				Provider:    "claude",
@@ -829,7 +830,7 @@ func TestExecutionService_AgentStep_MultipleProviders(t *testing.T) {
 	registry := mocks.NewMockAgentRegistry()
 
 	claude := mocks.NewMockAgentProvider("claude")
-	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any) (*workflow.AgentResult, error) {
+	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any, stdout, stderr io.Writer) (*workflow.AgentResult, error) {
 		if prompt == "Analyze this code" {
 			return &workflow.AgentResult{
 				Provider:    "claude",
@@ -852,7 +853,7 @@ func TestExecutionService_AgentStep_MultipleProviders(t *testing.T) {
 	_ = registry.Register(claude)
 
 	gemini := mocks.NewMockAgentProvider("gemini")
-	gemini.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any) (*workflow.AgentResult, error) {
+	gemini.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any, stdout, stderr io.Writer) (*workflow.AgentResult, error) {
 		if prompt == "Review the analysis" {
 			return &workflow.AgentResult{
 				Provider:    "gemini",
@@ -1004,7 +1005,7 @@ func TestExecutionService_Resume_WithAgentStep(t *testing.T) {
 
 	registry := mocks.NewMockAgentRegistry()
 	claude := mocks.NewMockAgentProvider("claude")
-	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any) (*workflow.AgentResult, error) {
+	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any, stdout, stderr io.Writer) (*workflow.AgentResult, error) {
 		if prompt == "Summarize this text" {
 			return &workflow.AgentResult{
 				Provider:    "claude",
@@ -1073,7 +1074,7 @@ func TestExecutionService_AgentStep_ContinueOnError(t *testing.T) {
 
 	registry := mocks.NewMockAgentRegistry()
 	claude := mocks.NewMockAgentProvider("claude")
-	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any) (*workflow.AgentResult, error) {
+	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any, stdout, stderr io.Writer) (*workflow.AgentResult, error) {
 		if prompt == "Summarize this text" {
 			return &workflow.AgentResult{
 				Provider:    "claude",
@@ -1252,7 +1253,7 @@ func TestExecutionService_AgentStep_ExecutionError(t *testing.T) {
 
 	registry := mocks.NewMockAgentRegistry()
 	claude := mocks.NewMockAgentProvider("claude")
-	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any) (*workflow.AgentResult, error) {
+	claude.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any, stdout, stderr io.Writer) (*workflow.AgentResult, error) {
 		return nil, errors.New("network connection failed")
 	})
 	_ = registry.Register(claude)
@@ -1393,7 +1394,7 @@ func TestExecutionService_AgentStep_ProviderInterpolation(t *testing.T) {
 			registry := mocks.NewMockAgentRegistry()
 			for _, name := range tt.registeredNames {
 				provider := mocks.NewMockAgentProvider(name)
-				provider.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any) (*workflow.AgentResult, error) {
+				provider.SetExecuteFunc(func(ctx context.Context, prompt string, options map[string]any, stdout, stderr io.Writer) (*workflow.AgentResult, error) {
 					return &workflow.AgentResult{
 						Provider:    name,
 						Output:      "Result from " + name,

internal/application/conversation_manager.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"errors"
 	"fmt"
+	"io"
 	"strings"
 
 	"github.com/awf-project/cli/internal/domain/ports"
@@ -118,14 +119,15 @@ func (m *ConversationManager) executeTurn(
 	state *workflow.ConversationState,
 	prompt string,
 	options map[string]any,
+	stdoutW, stderrW io.Writer,
 ) (*workflow.ConversationResult, error) {
 	select {
 	case <-ctx.Done():
 		return nil, ctx.Err()
 	default:
 	}
 
-	result, err := provider.ExecuteConversation(ctx, state, prompt, options)
+	result, err := provider.ExecuteConversation(ctx, state, prompt, options, stdoutW, stderrW)
 	if err != nil {
 		return nil, err
 	}
@@ -197,6 +199,7 @@ func (m *ConversationManager) ExecuteConversation(
 	config *workflow.ConversationConfig,
 	execCtx *workflow.ExecutionContext,
 	buildContext ContextBuilderFunc,
+	stdoutW, stderrW io.Writer,
 ) (*workflow.ConversationResult, error) {
 	if err := m.validateConversationInputs(step, config); err != nil {
 		return nil, err
@@ -233,7 +236,7 @@ func (m *ConversationManager) ExecuteConversation(
 
 	var lastResult *workflow.ConversationResult
 	for turnCount := 0; turnCount < maxTurns; turnCount++ {
-		result, err := m.executeTurn(ctx, provider, state, resolvedPrompt, options)
+		result, err := m.executeTurn(ctx, provider, state, resolvedPrompt, options, stdoutW, stderrW)
 		if err != nil {
 			return nil, err
 		}

internal/application/conversation_manager_helpers_test.go

@@ -3,6 +3,7 @@ package application
 import (
 	"context"
 	"errors"
+	"io"
 	"testing"
 
 	"github.com/awf-project/cli/internal/domain/ports"
@@ -29,7 +30,7 @@ func (m *mockAgentProvider) Name() string {
 	return m.name
 }
 
-func (m *mockAgentProvider) Execute(ctx context.Context, prompt string, options map[string]any) (*workflow.AgentResult, error) {
+func (m *mockAgentProvider) Execute(ctx context.Context, prompt string, options map[string]any, stdout, stderr io.Writer) (*workflow.AgentResult, error) {
 	return nil, nil // Not used in conversation manager tests
 }
 
@@ -38,6 +39,7 @@ func (m *mockAgentProvider) ExecuteConversation(
 	state *workflow.ConversationState,
 	prompt string,
 	options map[string]any,
+	stdout, stderr io.Writer,
 ) (*workflow.ConversationResult, error) {
 	if m.err != nil {
 		return nil, m.err
@@ -433,7 +435,7 @@ func TestConversationManager_executeTurn(t *testing.T) {
 				cancel() // Cancel immediately
 			}
 
-			result, err := mgr.executeTurn(ctx, provider, tt.state, tt.prompt, tt.options)
+			result, err := mgr.executeTurn(ctx, provider, tt.state, tt.prompt, tt.options, nil, nil)
 
 			if tt.expectError {
 				assert.Error(t, err)