fix: error handling, sandbox env, file tool enforcement

- tools/file_*: hard-fail when OLLIE_SESSION_ID is unset; surface errors
  on all failure paths including dir checks, marker cleanup, and I/O
- tools/file_read: treat nonexistent path as new file (emit "(new file)",
  write empty marker) so file_write can create new files
- tools/file_write: colorize new-file lines with ＋ (U+FF0B)
- tools/file_glob, file_grep, memory_*, task_*: consistent error handling
  and output to stdout so the model sees all failures
- tui/diffcolor.go: match ＋/− (U+FF0B/U+2212) lookalikes emitted by
  file_edit/file_write; fix +++ / --- header pattern
- prompts: prohibit execute_code for file creation/modification;
  enforce execute_tool with file_write/file_edit instead
- core submodule: surface isError from MCP responses; inject
  OLLIE_SESSION_ID into execute server at session creation

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

core

@@ -1 +1 @@
-Subproject commit 64d3915593c9600911cf332b9f0d42bc8a4e2d5d
+Subproject commit 655b9e706045f503b4889c6bee6e43261b50f4f0

prompts/00_base.md

@@ -1,5 +1,7 @@
 You are ollie, an agentic assistant.
 
+**Critical rule: never use `execute_code` to create or modify files. Always use `execute_tool` with `file_write` or `file_edit`. No exceptions.**
+
 # Tone and style
 
 - Format output as markdown. The frontend may render it as HTML, display it in a terminal, or show it as plain text — never rely on ANSI escape codes or terminal-specific formatting.
@@ -20,6 +22,7 @@ Be truthful. If the user is wrong, say so — do not soften, hedge, or dance aro
 # Tool preferences
 
 - Make independent tool calls in parallel when there are no dependencies between them.
+- **Never use `execute_code` to create or modify files.** Use `execute_tool` with `file_write` or `file_edit` instead. This is a hard rule — there are no exceptions for "simple" files.
 
 # Environment
 

prompts/07_edit-text.md

@@ -18,7 +18,7 @@ Using `cat`, `sed`, `awk`, `grep`, `find`, `echo >`, or similar shell commands d
 
 - **Read before writing.** Always call `file_read` on a file before `file_write` or `file_edit`. This is enforced — writes will fail without a prior read.
 - **Prefer `file_edit` over `file_write`** for partial changes.
-- **Never use `execute_code` or `execute_tool` for file I/O.** Reserve those for operations that genuinely require a shell: building, testing, git, process control.
+- **To create or modify a file with content you provide, always use `execute_tool` with `file_write` or `file_edit` — never `execute_code`.** Capturing process output to a file (e.g. `make > make.log`, `openssl genrsa > key.pem`) is fine.
 
 ## Examples
 

tools/file_edit

@@ -8,14 +8,27 @@ import hashlib
 import difflib
 
 def marker_path(file_path):
+    sid = os.environ.get('OLLIE_SESSION_ID', '')
+    if not sid:
+        print("error: OLLIE_SESSION_ID is not set")
+        sys.exit(1)
     h = hashlib.sha256(file_path.encode()).hexdigest()[:64]
-    sid = os.environ.get('OLLIE_SESSION_ID', '_')
     return f"/tmp/ollie/{sid}/{h}"
 
 def unified_diff(path, old, new):
     old_lines = old.splitlines(keepends=True)
     new_lines = new.splitlines(keepends=True)
-    return ''.join(difflib.unified_diff(old_lines, new_lines, fromfile=path, tofile=path))
+    raw = ''.join(difflib.unified_diff(old_lines, new_lines, fromfile=path, tofile=path))
+    # Replace leading +/- with Unicode lookalikes so only the colorizer's
+    # explicit patterns trigger, not incidental ASCII +/- in content.
+    out = []
+    for line in raw.splitlines(keepends=True):
+        if line.startswith('+'):
+            line = '＋' + line[1:]
+        elif line.startswith('-'):
+            line = '−' + line[1:]
+        out.append(line)
+    return ''.join(out)
 
 def find_exact(content, find):
     return find if find in content else None
@@ -124,7 +137,11 @@ except OSError as e:
     print(f"error: {e}")
     sys.exit(1)
 
-os.unlink(marker)
+try:
+    os.unlink(marker)
+except OSError as e:
+    print(f"error: failed to remove marker: {e}")
+    sys.exit(1)
 
 diff = unified_diff(file_path, original, new_content)
 print(diff if diff else "(no changes)", end='' if diff else '\n')

tools/file_glob

@@ -7,8 +7,23 @@ search_path="${2:-$PWD}"
 
 [ -z "$pattern" ] && echo "usage: file_glob <pattern> [path]" && exit 1
 
+if [ ! -d "$search_path" ]; then
+    echo "error: path is not a directory: $search_path"
+    exit 1
+fi
+
 # collect matches and sort by mtime descending
-rg --no-config --files -g "$pattern" "$search_path" 2>/dev/null \
-    | xargs -d '\n' stat --printf '%Y\t%n\n' 2>/dev/null \
+rg_out=$(rg --no-config --files -g "$pattern" "$search_path" 2>&1)
+rg_exit=$?
+if [ $rg_exit -gt 1 ]; then
+    echo "error: $rg_out"
+    exit 1
+fi
+if [ -z "$rg_out" ]; then
+    echo "no matches"
+    exit 0
+fi
+echo "$rg_out" \
+    | xargs -d '\n' stat --printf '%Y\t%n\n' 2>&1 \
     | sort -rn \
     | cut -f2-

tools/file_grep

@@ -57,9 +57,10 @@ cmd += ['-e', args.pattern, search_path]
 
 result = subprocess.run(cmd, capture_output=True, text=True)
 if result.returncode == 1:
-    sys.exit(0)  # no matches
+    print("no matches")
+    sys.exit(0)
 if result.returncode > 1:
-    print(result.stderr.strip() or f"rg exited with {result.returncode}", file=sys.stderr)
+    print(f"error: {result.stderr.strip() or f'rg exited with {result.returncode}'}")
     sys.exit(1)
 
 lines = result.stdout.splitlines()

tools/file_read

@@ -10,8 +10,21 @@ max_size=$((8 * 1024 * 1024))
 
 [ -z "$file_path" ] && echo "usage: file_read <file_path> [offset] [limit]" && exit 1
 [[ "$file_path" != /* ]] && echo "error: path must be absolute, got: $file_path" && exit 1
-[ ! -e "$file_path" ] && echo "error: file not found: $file_path" && exit 1
-[ ! -f "$file_path" ] && echo "error: not a regular file: $file_path" && exit 1
+[ ! -f "$file_path" ] && [ -e "$file_path" ] && echo "error: not a regular file: $file_path" && exit 1
+
+if [ -z "$OLLIE_SESSION_ID" ]; then
+    echo "error: OLLIE_SESSION_ID is not set"
+    exit 1
+fi
+
+mkdir -p "/tmp/ollie/$OLLIE_SESSION_ID"
+marker="/tmp/ollie/$OLLIE_SESSION_ID/$(printf '%s' "$file_path" | sha256sum | cut -c1-64)"
+
+if [ ! -e "$file_path" ]; then
+    echo "(new file)"
+    : > "$marker"
+    exit 0
+fi
 
 size=$(wc -c < "$file_path")
 [ "$size" -gt "$max_size" ] && echo "error: file too large ($size bytes, max $max_size)" && exit 1
@@ -31,6 +44,4 @@ else
 fi
 
 # save content to marker (used by file_edit/file_write as staging area)
-mkdir -p "/tmp/ollie/$OLLIE_SESSION_ID"
-marker="/tmp/ollie/$OLLIE_SESSION_ID/$(printf '%s' "$file_path" | sha256sum | cut -c1-64)"
 cp "$file_path" "$marker"

tools/file_write

@@ -8,8 +8,11 @@ import hashlib
 import difflib
 
 def marker_path(file_path):
+    sid = os.environ.get('OLLIE_SESSION_ID', '')
+    if not sid:
+        print("error: OLLIE_SESSION_ID is not set")
+        sys.exit(1)
     h = hashlib.sha256(file_path.encode()).hexdigest()[:64]
-    sid = os.environ.get('OLLIE_SESSION_ID', '_')
     return f"/tmp/ollie/{sid}/{h}"
 
 def unified_diff(path, old, new):
@@ -39,6 +42,11 @@ if exists:
     with open(marker, 'r', errors='replace') as f:
         old_content = f.read()
 
+parent = os.path.dirname(file_path)
+if parent and not os.path.isdir(parent):
+    print(f"error: parent directory does not exist: {parent}")
+    sys.exit(1)
+
 try:
     with open(marker, 'w') as f:
         f.write(content)
@@ -48,11 +56,15 @@ except OSError as e:
     print(f"error: {e}")
     sys.exit(1)
 
-os.unlink(marker)
+try:
+    os.unlink(marker)
+except OSError as e:
+    print(f"error: failed to remove marker: {e}")
+    sys.exit(1)
 
 if not exists:
     for line in content.splitlines():
-        print(f"+{line}")
+        print(f"＋{line}")
 else:
     diff = unified_diff(file_path, old_content, content)
     print(diff if diff else "(no changes)", end='' if diff else '\n')

tools/memory_recall

@@ -7,10 +7,20 @@
 mount="${OLLIE:-$HOME/mnt/ollie}"
 mem_dir="$mount/m"
 
+if [ ! -d "$mem_dir" ]; then
+    echo "error: memory directory does not exist: $mem_dir"
+    exit 1
+fi
+
 query="$*"
 
 # Search filenames (title/tag match) and file bodies
-matches=$(grep -ril "$query" "$mem_dir" 2>/dev/null)
+matches=$(grep -ril "$query" "$mem_dir" 2>&1)
+grep_exit=$?
+if [ $grep_exit -gt 1 ]; then
+    echo "error: $matches"
+    exit 1
+fi
 
 if [ -z "$matches" ]; then
     echo "no matches for: $query"
@@ -19,6 +29,6 @@ fi
 
 while IFS= read -r file; do
     echo "=== $(basename "$file") ==="
-    cat "$file"
+    cat "$file" || echo "error: could not read $file"
     echo
 done <<< "$matches"

tools/memory_remember

@@ -18,6 +18,10 @@ body     = sys.argv[3]
 mount = os.environ.get("OLLIE") or os.path.expanduser("~/mnt/ollie")
 mem_dir = os.path.join(mount, "m")
 
+if not os.path.isdir(mem_dir):
+    print(f"error: memory directory does not exist: {mem_dir}")
+    sys.exit(1)
+
 now   = datetime.now()
 ident = now.strftime("%Y%m%dT%H%M%S")
 date  = now.strftime("%Y-%m-%d %a %H:%M")

tools/task_complete

@@ -4,7 +4,7 @@
 import sys, os, re
 
 if len(sys.argv) < 3:
-    print("usage: task_complete <plan-file> <step-substring>", file=sys.stderr)
+    print("usage: task_complete <plan-file> <step-substring>", file=sys.stdout)
     sys.exit(1)
 
 ollie = os.environ.get('OLLIE', os.path.expanduser('~/mnt/ollie'))
@@ -14,8 +14,16 @@ plan_dir = os.path.join(ollie, 'pl')
 filepath = os.path.join(plan_dir, os.path.basename(sys.argv[1]))
 substring = sys.argv[2].lower()
 
-with open(filepath) as f:
-    lines = f.readlines()
+if not os.path.exists(filepath):
+    print(f"error: plan file not found: {filepath}", file=sys.stdout)
+    sys.exit(1)
+
+try:
+    with open(filepath) as f:
+        lines = f.readlines()
+except OSError as e:
+    print(f"error: {e}", file=sys.stdout)
+    sys.exit(1)
 
 matched = False
 for i, line in enumerate(lines):
@@ -25,7 +33,7 @@ for i, line in enumerate(lines):
         break
 
 if not matched:
-    print(f"no unchecked step matching {sys.argv[2]!r}", file=sys.stderr)
+    print(f"no unchecked step matching {sys.argv[2]!r}", file=sys.stdout)
     sys.exit(1)
 
 remaining = [line[6:].strip() for line in lines if re.match(r'^- \[ \] ', line)]
@@ -40,11 +48,19 @@ elif not remaining and '__wip' in filepath:
 else:
     newpath = filepath
 
-with open(filepath, 'w') as f:
-    f.writelines(lines)
+try:
+    with open(filepath, 'w') as f:
+        f.writelines(lines)
+except OSError as e:
+    print(f"error: failed to write {filepath}: {e}", file=sys.stdout)
+    sys.exit(1)
 
 if newpath != filepath:
-    os.rename(filepath, newpath)
+    try:
+        os.rename(filepath, newpath)
+    except OSError as e:
+        print(f"error: failed to rename to {newpath}: {e}", file=sys.stdout)
+        sys.exit(1)
 
 if remaining:
     print('\n'.join(f'- {s}' for s in remaining))

tools/task_plan

@@ -8,14 +8,19 @@ OLLIE="${OLLIE:-$HOME/mnt/ollie}"
 PLAN_DIR="$OLLIE/pl"
 
 if [ $# -eq 0 ]; then
-    echo "usage: task_plan <goal> <step1> [step2] ..." >&2
+    echo "usage: task_plan <goal> <step1> [step2] ..."
     exit 1
 fi
 GOAL="$1"
 shift
 
 if [ $# -eq 0 ]; then
-    echo "error: at least one step required" >&2
+    echo "error: at least one step required"
+    exit 1
+fi
+
+if [ ! -d "$PLAN_DIR" ]; then
+    echo "error: plan directory does not exist: $PLAN_DIR"
     exit 1
 fi
 
@@ -36,6 +41,6 @@ FILEPATH="$PLAN_DIR/$FILENAME"
     for step in "$@"; do
         printf '- [ ] %s\n' "$step"
     done
-} > "$FILEPATH"
+} > "$FILEPATH" || { echo "error: failed to write $FILEPATH"; exit 1; }
 
 echo "$FILENAME"