Add loop tests (self-generated) + --agent flag for all commands

.github/workflows/ci.yml

@@ -19,6 +19,7 @@ jobs:
       - uses: actions/setup-go@v5
         with:
           go-version-file: src/go.mod
+          cache-dependency-path: src/go.sum
 
       - name: Build
         run: go build ./...

.github/workflows/release.yml

@@ -62,6 +62,10 @@ jobs:
           GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
         run: |
           TAG="v${{ steps.bump.outputs.version }}"
+          if git ls-remote --tags origin "refs/tags/$TAG" | grep -q "$TAG"; then
+            echo "Tag $TAG already exists — skipping release"
+            exit 0
+          fi
           git tag "$TAG"
           git push origin "$TAG"
           gh release create "$TAG" --title "$TAG" --notes "${{ steps.notes.outputs.notes }}"

README.md

@@ -56,6 +56,7 @@ Works with just Claude. Optionally adds Codex and Gemini for multi-perspective a
 | `/devkit:refactor` | Full refactor lifecycle — analyze, plan, restructure, verify, compare |
 | `/devkit:research` | Deep research — clarify, parallel search, analyze sources, synthesize |
 | `/devkit:decompose` | Goal decomposition — break into task DAG, assign agents, execute in dependency order |
+| `/devkit:scrape` | URL-to-Markdown conversion via Jina Reader / Firecrawl / WebFetch |
 | `/devkit:status` | Health check — installed CLIs, available agents, ready commands |
 
 ### Self-Improvement Loops (Claude-only)
@@ -268,8 +269,17 @@ devkit/
 │   └── tri-security.yml     # Three-tier security audit
 ├── presets/                  # Reusable prompt templates (planned)
 │   └── .gitkeep
-└── src/
-    └── TODO.md              # Go harness roadmap
+├── .github/workflows/       # CI/CD
+│   ├── ci.yml               # Build + test + vet on push/PR
+│   └── release.yml          # Auto-tag + release on version bump
+└── src/                     # Go CLI harness
+    ├── go.mod
+    ├── main.go
+    ├── Makefile
+    ├── cmd/                 # Cobra commands
+    ├── lib/                 # DB, git, metric, state, report
+    ├── loops/               # Improve, feature, bugfix, refactor, testgen, review, dispatch
+    └── runners/             # Claude, Codex, Gemini runner interfaces
 ```
 
 ---
@@ -288,23 +298,31 @@ Set automatically in each multi-agent command:
 
 ## Go CLI Harness
 
-Deterministic orchestration binary — the machine controls the loop, Claude is the body.
+Deterministic orchestration binary — the machine controls the loop, the agent is the body.
 
 ```bash
-cd src && make build
-./bin/devkit --help
+cd src && make build && make link
+devkit --help
+```
+
+Or install directly:
+
+```bash
+cd src && make install
 ```
 
 ### Commands
 
+All loop commands support `--agent` to choose the AI agent (default: `claude`).
+
 | Command | Description |
 |---|---|
-| `devkit improve` | Metric-gated iteration loop — one Claude invocation per iteration |
+| `devkit improve` | Metric-gated iteration loop — one agent invocation per iteration |
 | `devkit feature` | Plan, implement, test, lint — commits only after tests pass |
 | `devkit bugfix` | Diagnose, fix, verify — reverts if tests break |
 | `devkit refactor` | Analyze, transform, verify — reverts if behavior changes |
 | `devkit test-gen` | Generate tests, run, fix failures — iterates until green |
-| `devkit review` | Parallel multi-agent code review (Claude + Codex + Gemini) |
+| `devkit review` | Parallel multi-agent code review |
 | `devkit dispatch` | Send any task to multiple agents, compare outputs |
 | `devkit status` | Show all sessions, costs, iteration history |
 | `devkit resume` | Pick up a crashed or paused session |
@@ -316,13 +334,17 @@ cd src && make build
 - **Hard budget caps** — stops spawning at your dollar limit
 - **CI/CD integration** — runs headless, no conversation needed
 - **True parallel dispatch** — goroutines, not sequential prompts
+- **Multi-agent support** — `--agent claude`, `--agent codex`, or `--agent gemini`
 
 ### Examples
 
 ```bash
 # Run 50 improvement iterations overnight, stop at $20
 devkit improve --metric "npm test" --iterations 50 --budget 20.00
 
+# Same thing with Codex instead of Claude
+devkit improve --metric "npm test" --iterations 50 --agent codex
+
 # Implement a feature with test verification
 devkit feature "add JWT auth" --target src/auth/ --test "npm test"
 
@@ -332,13 +354,29 @@ devkit bugfix "login 500 on plus sign emails" --test "go test ./..."
 # Generate tests for a module
 devkit test-gen src/parser/ --test "go test ./..."
 
+# Multi-agent review with all available agents
+devkit review
+
 # Resume a crashed session
 devkit resume abc123def456
 
 # Check what happened
 devkit status
 ```
 
+### Testing
+
+```bash
+cd src && go test ./... -v
+```
+
+76+ tests across 4 packages (lib, runners, loops, cmd). Loop tests use mock runners — no API calls needed.
+
+### CI/CD
+
+- **CI pipeline** (`.github/workflows/ci.yml`): build + vet + test (with `-race`) + gofmt check on every push/PR to main
+- **Auto-release** (`.github/workflows/release.yml`): auto-bumps version, tags, and creates GitHub release on merged PRs
+
 ---
 
 ## Prerequisites
@@ -404,7 +442,9 @@ Devkit's `tri:*` commands use `/gemini:rescue --background` for multi-agent disp
 
 See [ROADMAP.md](ROADMAP.md) for full details.
 
-- [ ] Go CLI harness for deterministic loop control, process management, and unattended runs (see `src/TODO.md`)
+- [x] Go CLI harness — 9 commands, SQLite state, crash recovery, budget enforcement, multi-agent support
+- [x] CI/CD pipeline — build, vet, test, auto-release on version bump
+- [x] Branch protection — PRs required for main
 - [ ] Cost event hooks — budget threshold events with auto-downgrade actions
 - [ ] Execution registry — centralized step tracking with timing and token usage
 - [ ] Preset library — curated prompt templates for common review/improvement scenarios

src/cmd/bugfix.go

@@ -6,14 +6,13 @@ import (
 
 	"github.com/5uck1ess/devkit/lib"
 	"github.com/5uck1ess/devkit/loops"
-	"github.com/5uck1ess/devkit/runners"
 	"github.com/spf13/cobra"
 )
 
 var bugfixCmd = &cobra.Command{
 	Use:   "bugfix [description]",
 	Short: "Full bugfix lifecycle: diagnose, fix, verify",
-	Long:  "Spawns Claude for each step: diagnose root cause → apply fix → run tests to verify.",
+	Long:  "Spawns an AI agent for each step: diagnose root cause → apply fix → run tests to verify.",
 	Example: `  devkit bugfix "login returns 500 when email has a plus sign"
   devkit bugfix "race condition in cache invalidation" --test "go test ./..."`,
 	Args: cobra.MinimumNArgs(1),
@@ -29,10 +28,10 @@ var bugfixCmd = &cobra.Command{
 			return fmt.Errorf("working tree has uncommitted changes — commit or stash first")
 		}
 
-		available := runners.DetectRunners()
-		runner := runners.FindRunner("claude", available)
-		if runner == nil {
-			return fmt.Errorf("claude CLI not found in PATH")
+		agentName, _ := cmd.Flags().GetString("agent")
+		runner, err := resolveRunner(agentName)
+		if err != nil {
+			return err
 		}
 
 		result, err := loops.RunBugfix(cmd.Context(), db, runner, &lib.Git{Dir: repoRoot}, loops.BugfixConfig{

src/cmd/feature.go

@@ -6,14 +6,13 @@ import (
 
 	"github.com/5uck1ess/devkit/lib"
 	"github.com/5uck1ess/devkit/loops"
-	"github.com/5uck1ess/devkit/runners"
 	"github.com/spf13/cobra"
 )
 
 var featureCmd = &cobra.Command{
 	Use:   "feature [description]",
 	Short: "Full feature lifecycle: plan, implement, test, lint",
-	Long:  "Spawns Claude for each step: plan → implement → test (loop until green) → lint.",
+	Long:  "Spawns an AI agent for each step: plan → implement → test (loop until green) → lint.",
 	Example: `  devkit feature "add JWT authentication" --target src/auth/
   devkit feature "add search endpoint" --test "npm test" --lint "npm run lint"`,
 	Args: cobra.MinimumNArgs(1),
@@ -31,10 +30,10 @@ var featureCmd = &cobra.Command{
 			return fmt.Errorf("working tree has uncommitted changes — commit or stash first")
 		}
 
-		available := runners.DetectRunners()
-		runner := runners.FindRunner("claude", available)
-		if runner == nil {
-			return fmt.Errorf("claude CLI not found in PATH")
+		agentName, _ := cmd.Flags().GetString("agent")
+		runner, err := resolveRunner(agentName)
+		if err != nil {
+			return err
 		}
 
 		result, err := loops.RunFeature(cmd.Context(), db, runner, &lib.Git{Dir: repoRoot}, loops.FeatureConfig{

src/cmd/improve.go

@@ -5,14 +5,13 @@ import (
 
 	"github.com/5uck1ess/devkit/lib"
 	"github.com/5uck1ess/devkit/loops"
-	"github.com/5uck1ess/devkit/runners"
 	"github.com/spf13/cobra"
 )
 
 var improveCmd = &cobra.Command{
 	Use:   "improve",
 	Short: "Run a metric-gated improvement loop",
-	Long:  "Spawns Claude per iteration. Each iteration: propose change, run metric, keep if pass, revert if fail.",
+	Long:  "Spawns an AI agent per iteration. Each iteration: propose change, run metric, keep if pass, revert if fail.",
 	Example: `  devkit improve --target src/ --metric "npm test" --objective "0 failing tests" --iterations 20
   devkit improve --metric "go test ./..." --iterations 10 --budget 5.00`,
 	RunE: func(cmd *cobra.Command, args []string) error {
@@ -43,10 +42,10 @@ var improveCmd = &cobra.Command{
 			return fmt.Errorf("working tree has uncommitted changes — commit or stash before running devkit improve")
 		}
 
-		available := runners.DetectRunners()
-		runner := runners.FindRunner("claude", available)
-		if runner == nil {
-			return fmt.Errorf("claude CLI not found in PATH — install from https://docs.anthropic.com/en/docs/claude-code")
+		agentName, _ := cmd.Flags().GetString("agent")
+		runner, err := resolveRunner(agentName)
+		if err != nil {
+			return err
 		}
 
 		git := &lib.Git{Dir: repoRoot}

src/cmd/refactor.go

@@ -6,14 +6,13 @@ import (
 
 	"github.com/5uck1ess/devkit/lib"
 	"github.com/5uck1ess/devkit/loops"
-	"github.com/5uck1ess/devkit/runners"
 	"github.com/spf13/cobra"
 )
 
 var refactorCmd = &cobra.Command{
 	Use:   "refactor [description]",
 	Short: "Full refactor lifecycle: analyze, transform, verify",
-	Long:  "Spawns Claude for each step: analyze code smells → apply transformations → verify tests still pass.",
+	Long:  "Spawns an AI agent for each step: analyze code smells → apply transformations → verify tests still pass.",
 	Example: `  devkit refactor "extract auth middleware into shared package" --target src/
   devkit refactor "flatten nested callbacks to async/await" --test "npm test"`,
 	Args: cobra.MinimumNArgs(1),
@@ -30,10 +29,10 @@ var refactorCmd = &cobra.Command{
 			return fmt.Errorf("working tree has uncommitted changes — commit or stash first")
 		}
 
-		available := runners.DetectRunners()
-		runner := runners.FindRunner("claude", available)
-		if runner == nil {
-			return fmt.Errorf("claude CLI not found in PATH")
+		agentName, _ := cmd.Flags().GetString("agent")
+		runner, err := resolveRunner(agentName)
+		if err != nil {
+			return err
 		}
 
 		result, err := loops.RunRefactor(cmd.Context(), db, runner, &lib.Git{Dir: repoRoot}, loops.RefactorConfig{

src/cmd/resume.go

@@ -7,7 +7,6 @@ import (
 
 	"github.com/5uck1ess/devkit/lib"
 	"github.com/5uck1ess/devkit/loops"
-	"github.com/5uck1ess/devkit/runners"
 	"github.com/spf13/cobra"
 )
 
@@ -40,10 +39,10 @@ var resumeCmd = &cobra.Command{
 			return fmt.Errorf("resume only supports improve sessions — session %s is %q", sessionID, session.Workflow)
 		}
 
-		available := runners.DetectRunners()
-		runner := runners.FindRunner("claude", available)
-		if runner == nil {
-			return fmt.Errorf("claude CLI not found in PATH")
+		agentName, _ := cmd.Flags().GetString("agent")
+		runner, err := resolveRunner(agentName)
+		if err != nil {
+			return err
 		}
 
 		// Require clean worktree before resuming

src/cmd/root.go

@@ -6,9 +6,11 @@ import (
 	"os"
 	"os/signal"
 	"path/filepath"
+	"strings"
 	"syscall"
 
 	"github.com/5uck1ess/devkit/lib"
+	"github.com/5uck1ess/devkit/runners"
 	"github.com/spf13/cobra"
 )
 
@@ -44,6 +46,10 @@ var rootCmd = &cobra.Command{
 	},
 }
 
+func init() {
+	rootCmd.PersistentFlags().String("agent", "claude", "AI agent to use (claude, codex, gemini)")
+}
+
 func Execute() error {
 	rootCmd.Version = Version
 	ctx, cancel := context.WithCancel(context.Background())
@@ -60,6 +66,26 @@ func Execute() error {
 	return rootCmd.ExecuteContext(ctx)
 }
 
+func resolveRunner(name string) (runners.Runner, error) {
+	return resolveRunnerFrom(name, runners.DetectRunners())
+}
+
+func resolveRunnerFrom(name string, available []runners.Runner) (runners.Runner, error) {
+	name = strings.ToLower(strings.TrimSpace(name))
+	r := runners.FindRunner(name, available)
+	if r != nil {
+		return r, nil
+	}
+	var names []string
+	for _, a := range available {
+		names = append(names, a.Name())
+	}
+	if len(names) == 0 {
+		return nil, fmt.Errorf("no AI agents found in PATH — install claude, codex, or gemini")
+	}
+	return nil, fmt.Errorf("agent %q not found — available: %s", name, strings.Join(names, ", "))
+}
+
 func findRepoRoot() (string, error) {
 	dir, err := os.Getwd()
 	if err != nil {