Screenshots from implementers

internal/agents/hands.go

@@ -2,6 +2,12 @@ package agents
 
 import "g.fc.run/misc/byre/internal/mcp"
 
-func handsMCP() mcp.Stdio {
-	return mcp.Stdio{Name: "hands", Command: "byre-hands", Args: []string{}}
+const screenshotGuidance = "\n- When a turn changes something you can see, you can show it in your final message: save screenshots with the Playwright tools (pages you serve on this task's ports) or hands' capture with save (apps on this Mac), and embed each as a Markdown image, ![what it shows](<path>), with the path the tool returned. They go to the task's media folder, outside the worktree, so they are never committed, and are deleted when the task closes. Leave out a screenshot that adds nothing to the text."
+
+func MediaDir(worktreePath string) string {
+	return worktreePath + "-media"
+}
+
+func handsMCP(mediaDir string) mcp.Stdio {
+	return mcp.Stdio{Name: "hands", Command: "byre-hands", Args: []string{"--media-dir", mediaDir}}
 }

internal/agents/manager.go

@@ -734,7 +734,13 @@ func (m *Manager) startLive(ctx context.Context, sess api.AgentSession, task api
 	env := portEnv(task)
 	stdio := []mcp.Stdio{}
 	if sess.Role == api.Implementer {
-		stdio = []mcp.Stdio{handsMCP()}
+		media := MediaDir(task.WorktreePath)
+		err = os.MkdirAll(media, 0o755)
+		if err != nil {
+			return nil, err
+		}
+		g += screenshotGuidance
+		stdio = []mcp.Stdio{handsMCP(media), playwrightMCP(media)}
 	}
 	if sess.Role == api.Reviewer {
 		r, err := m.store.ReviewBySession(ctx, sess.Id)

internal/agents/mergein.go

@@ -6,6 +6,7 @@ import (
 	"errors"
 	"fmt"
 	"log/slog"
+	"os"
 	"slices"
 	"strings"
 	"sync"
@@ -572,6 +573,10 @@ func (m *Manager) RemoveTask(ctx context.Context, task api.Task, project api.Pro
 	if err != nil {
 		return err
 	}
+	err = os.RemoveAll(MediaDir(task.WorktreePath))
+	if err != nil {
+		return err
+	}
 	return git.RemoveWorktree(project.RepoPath, task.WorktreePath, task.Branch)
 }
 

internal/server/images.go

@@ -9,6 +9,7 @@ import (
 	"path/filepath"
 	"strings"
 
+	"g.fc.run/misc/byre/internal/agents"
 	"g.fc.run/misc/byre/internal/api"
 )
 
@@ -37,6 +38,11 @@ func (s *Server) GetTaskImage(w http.ResponseWriter, r *http.Request, taskID str
 		writeError(w, http.StatusInternalServerError, err)
 		return
 	}
+	media, err := filepath.EvalSymlinks(agents.MediaDir(t.WorktreePath))
+	if err != nil && !errors.Is(err, fs.ErrNotExist) {
+		writeError(w, http.StatusInternalServerError, err)
+		return
+	}
 
 	path := params.Path
 	if !filepath.IsAbs(path) {
@@ -52,9 +58,8 @@ func (s *Server) GetTaskImage(w http.ResponseWriter, r *http.Request, taskID str
 		return
 	}
 
-	rel, err := filepath.Rel(root, real)
-	if err != nil || rel == ".." || strings.HasPrefix(rel, ".."+string(filepath.Separator)) {
-		writeError(w, http.StatusForbidden, fmt.Errorf("%s is outside the task's worktree", params.Path))
+	if !within(root, real) && (media == "" || !within(media, real)) {
+		writeError(w, http.StatusForbidden, fmt.Errorf("%s is outside the task's worktree and media folder", params.Path))
 		return
 	}
 
@@ -87,3 +92,8 @@ func (s *Server) GetTaskImage(w http.ResponseWriter, r *http.Request, taskID str
 	w.Header().Set("Cache-Control", "private, no-cache")
 	http.ServeContent(w, r, "", info.ModTime(), f)
 }
+
+func within(root, path string) bool {
+	rel, err := filepath.Rel(root, path)
+	return err == nil && rel != ".." && !strings.HasPrefix(rel, ".."+string(filepath.Separator))
+}