Add tests, test-gen command, review fixes, README, CI

.github/workflows/ci.yml

@@ -0,0 +1,39 @@
+name: CI
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+    branches: [main]
+
+jobs:
+  build-and-test:
+    runs-on: ubuntu-latest
+    defaults:
+      run:
+        working-directory: src
+
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/setup-go@v5
+        with:
+          go-version-file: src/go.mod
+
+      - name: Build
+        run: go build ./...
+
+      - name: Vet
+        run: go vet ./...
+
+      - name: Test
+        run: go test ./... -v -race -count=1
+
+      - name: Check formatting
+        run: |
+          unformatted=$(gofmt -l .)
+          if [ -n "$unformatted" ]; then
+            echo "Files not formatted:"
+            echo "$unformatted"
+            exit 1
+          fi

README.md

@@ -286,6 +286,61 @@ Set automatically in each multi-agent command:
 
 ---
 
+## Go CLI Harness
+
+Deterministic orchestration binary — the machine controls the loop, Claude is the body.
+
+```bash
+cd src && make build
+./bin/devkit --help
+```
+
+### Commands
+
+| Command | Description |
+|---|---|
+| `devkit improve` | Metric-gated iteration loop — one Claude 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 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 |
+
+### What it does that plugins can't
+
+- **Exact iteration counts** — Go binary owns the loop, not the LLM
+- **Crash recovery** — SQLite state + handoff files survive crashes
+- **Hard budget caps** — stops spawning at your dollar limit
+- **CI/CD integration** — runs headless, no conversation needed
+- **True parallel dispatch** — goroutines, not sequential prompts
+
+### Examples
+
+```bash
+# Run 50 improvement iterations overnight, stop at $20
+devkit improve --metric "npm test" --iterations 50 --budget 20.00
+
+# Implement a feature with test verification
+devkit feature "add JWT auth" --target src/auth/ --test "npm test"
+
+# Fix a bug with automated verification
+devkit bugfix "login 500 on plus sign emails" --test "go test ./..."
+
+# Generate tests for a module
+devkit test-gen src/parser/ --test "go test ./..."
+
+# Resume a crashed session
+devkit resume abc123def456
+
+# Check what happened
+devkit status
+```
+
+---
+
 ## Prerequisites
 
 **Required:** Claude Code (you're already here)

commands/tri-debug.md

@@ -7,6 +7,18 @@ description: Multi-agent debugging — send a bug report to available agents (Cl
 
 Send a bug description to all available agents in parallel, get independent root-cause analyses, and consolidate into a recommended fix.
 
+## Step 0: Harness Detection
+
+```bash
+if command -v devkit >/dev/null 2>&1; then
+  echo "Go harness detected — delegating to devkit dispatch for full output capture."
+  devkit dispatch {prompt with bug context}
+  exit 0
+fi
+```
+
+If the `devkit` binary is in PATH, delegate entirely to it. Only fall through to plugin-based steps if the harness is not installed.
+
 ## Step 1: Gather Context
 
 Collect from the user:

commands/tri-dispatch.md

@@ -7,6 +7,18 @@ description: Dispatch a task to all three agents (Claude, Codex, Gemini) in para
 
 Send the same task to Claude, Codex, and Gemini in parallel. Compare outputs.
 
+## Step 0: Harness Detection
+
+```bash
+if command -v devkit >/dev/null 2>&1; then
+  echo "Go harness detected — delegating to devkit dispatch for full output capture."
+  devkit dispatch {prompt}
+  exit 0
+fi
+```
+
+If the `devkit` binary is in PATH, delegate entirely to it. Only fall through to plugin-based steps if the harness is not installed.
+
 ## When to use
 
 - Comparing approaches to a problem

commands/tri-review.md

@@ -7,6 +7,20 @@ description: Triple-agent PR/code review. Claude runs as native background agent
 
 Run the same code review across three AI agents in parallel and consolidate results.
 
+## Step 0: Harness Detection
+
+```bash
+if command -v devkit >/dev/null 2>&1; then
+  echo "Go harness detected — delegating to devkit review for full output capture."
+  devkit review {prompt or default}
+  # The harness handles parallel dispatch, full stdout capture (no truncation),
+  # SQLite session tracking, and consolidated output. Skip all steps below.
+  exit 0
+fi
+```
+
+If the `devkit` binary is in PATH, delegate entirely to it. The harness avoids output truncation, captures full agent responses, and tracks sessions in SQLite. Only fall through to the plugin-based steps below if the harness is not installed.
+
 ## Step 1: Gather Context
 
 ```bash

commands/tri-security.md

@@ -7,6 +7,18 @@ description: Multi-agent security audit — independent security reviews from av
 
 Independent security reviews from all available agents, consolidated into a severity-ranked report.
 
+## Step 0: Harness Detection
+
+```bash
+if command -v devkit >/dev/null 2>&1; then
+  echo "Go harness detected — delegating to devkit review --security for full output capture."
+  devkit review --security {prompt or default}
+  exit 0
+fi
+```
+
+If the `devkit` binary is in PATH, delegate entirely to it. Only fall through to plugin-based steps if the harness is not installed.
+
 ## Step 1: Gather Scope
 
 Determine what to audit:

commands/tri-test-gen.md

@@ -7,6 +7,18 @@ description: Multi-agent test generation — each available agent generates test
 
 Generate tests from all available agents in parallel, then merge the best tests into a comprehensive suite.
 
+## Step 0: Harness Detection
+
+```bash
+if command -v devkit >/dev/null 2>&1; then
+  echo "Go harness detected — delegating to devkit test-gen for full output capture."
+  devkit test-gen {target} --test {test_command}
+  exit 0
+fi
+```
+
+If the `devkit` binary is in PATH, delegate entirely to it. Only fall through to plugin-based steps if the harness is not installed.
+
 ## Step 1: Analyze Target
 
 Read the target files and detect:

src/Makefile

@@ -3,7 +3,7 @@ VERSION := $(shell git describe --tags --always --dirty 2>/dev/null || echo "dev
 LDFLAGS := -s -w -X main.version=$(VERSION)
 GOFLAGS := -trimpath
 
-.PHONY: build install clean test vet fmt check all
+.PHONY: build install link clean test vet fmt check all
 
 all: check build
 
@@ -12,6 +12,12 @@ build:
 
 install:
 	go install $(GOFLAGS) -ldflags '$(LDFLAGS)' .
+	@echo "Installed to $$(go env GOPATH)/bin/$(BINARY)"
+	@echo "Ensure $$(go env GOPATH)/bin is in your PATH"
+
+link: build
+	ln -sf $(CURDIR)/bin/$(BINARY) /usr/local/bin/$(BINARY)
+	@echo "Linked bin/$(BINARY) → /usr/local/bin/$(BINARY)"
 
 clean:
 	rm -rf bin/

src/cmd/dispatch.go

@@ -23,7 +23,12 @@ var dispatchCmd = &cobra.Command{
 
 		var agents []string
 		if agentList != "" {
-			agents = strings.Split(agentList, ",")
+			for _, a := range strings.Split(agentList, ",") {
+				a = strings.TrimSpace(a)
+				if a != "" {
+					agents = append(agents, a)
+				}
+			}
 		}
 
 		available := runners.DetectRunners()

src/cmd/resume.go

@@ -2,20 +2,27 @@ package cmd
 
 import (
 	"fmt"
+	"regexp"
+	"strings"
 
 	"github.com/5uck1ess/devkit/lib"
 	"github.com/5uck1ess/devkit/loops"
 	"github.com/5uck1ess/devkit/runners"
 	"github.com/spf13/cobra"
 )
 
+var sessionIDPattern = regexp.MustCompile(`^[a-f0-9]{12}$`)
+
 var resumeCmd = &cobra.Command{
 	Use:   "resume <session-id>",
 	Short: "Resume a paused or crashed session",
 	Long:  "Picks up an improve session from where it left off, using the SQLite state and handoff file.",
 	Args:  cobra.ExactArgs(1),
 	RunE: func(cmd *cobra.Command, args []string) error {
-		sessionID := args[0]
+		sessionID := strings.ToLower(args[0])
+		if !sessionIDPattern.MatchString(sessionID) {
+			return fmt.Errorf("invalid session ID %q — expected 12 hex characters (e.g., a1b2c3d4e5f6)", sessionID)
+		}
 
 		session, err := db.GetSession(sessionID)
 		if err != nil {

src/cmd/review.go

@@ -26,7 +26,12 @@ var reviewCmd = &cobra.Command{
 
 		var agents []string
 		if agentList != "" {
-			agents = strings.Split(agentList, ",")
+			for _, a := range strings.Split(agentList, ",") {
+				a = strings.TrimSpace(a)
+				if a != "" {
+					agents = append(agents, a)
+				}
+			}
 		}
 
 		available := runners.DetectRunners()

src/cmd/status.go

@@ -39,9 +39,13 @@ func showAllSessions() error {
 	fmt.Printf("%-14s %-10s %-10s %-8s %s\n", "-------", "--------", "------", "----", "-------")
 
 	for _, s := range sessions {
-		cost, _ := db.SessionTotalCost(s.ID)
+		cost, err := db.SessionTotalCost(s.ID)
+		costStr := fmt.Sprintf("$%.4f", cost)
+		if err != nil {
+			costStr = "unknown"
+		}
 		age := formatAge(s.CreatedAt)
-		fmt.Printf("%-14s %-10s %-10s $%-7.4f %s\n", s.ID, s.Workflow, s.Status, cost, age)
+		fmt.Printf("%-14s %-10s %-10s %-8s %s\n", s.ID, s.Workflow, s.Status, costStr, age)
 	}
 	return nil
 }

src/cmd/testgen.go

@@ -0,0 +1,66 @@
+package cmd
+
+import (
+	"fmt"
+	"strings"
+
+	"github.com/5uck1ess/devkit/lib"
+	"github.com/5uck1ess/devkit/loops"
+	"github.com/5uck1ess/devkit/runners"
+	"github.com/spf13/cobra"
+)
+
+var testGenCmd = &cobra.Command{
+	Use:   "test-gen [target]",
+	Short: "Generate tests for target code, run them, fix failures",
+	Long:  "Analyzes target code, generates comprehensive tests, runs them, and iterates until green.",
+	Example: `  devkit test-gen src/auth/
+  devkit test-gen lib/parser.go --test "go test ./..."
+  devkit test-gen src/ --test "npm test" --budget 5.00`,
+	Args: cobra.MinimumNArgs(1),
+	RunE: func(cmd *cobra.Command, args []string) error {
+		testCmd, _ := cmd.Flags().GetString("test")
+		budget, _ := cmd.Flags().GetFloat64("budget")
+
+		dirty, err := (&lib.Git{Dir: repoRoot}).HasUncommittedChanges()
+		if err != nil {
+			return fmt.Errorf("check git status: %w", err)
+		}
+		if dirty {
+			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")
+		}
+
+		result, err := loops.RunTestGen(cmd.Context(), db, runner, &lib.Git{Dir: repoRoot}, loops.TestGenConfig{
+			Target:   strings.Join(args, " "),
+			TestCmd:  testCmd,
+			RepoRoot: repoRoot,
+			BudgetUSD: budget,
+		})
+		if err != nil {
+			return err
+		}
+
+		var totalCost float64
+		for _, s := range result.Steps {
+			totalCost += s.CostUSD
+		}
+		fmt.Printf("\n=== Test Generation Complete ===\n")
+		fmt.Printf("Session: %s\n", result.Session.ID)
+		fmt.Printf("Steps:   %d\n", len(result.Steps))
+		fmt.Printf("Cost:    $%.4f\n", totalCost)
+		fmt.Printf("\nRun `devkit status %s` for details.\n", result.Session.ID)
+		return nil
+	},
+}
+
+func init() {
+	rootCmd.AddCommand(testGenCmd)
+	testGenCmd.Flags().String("test", "", "Test command to run generated tests")
+	testGenCmd.Flags().Float64("budget", 0, "Maximum spend in USD (0 = unlimited)")
+}