Handle directory inputs in to-markdown workflow

.devcontainer/opencode-seed/commands/to-markdown.md

@@ -9,18 +9,21 @@ Convert source content into Markdown using MarkItDown.
 
 ## Usage
 
-`/to-markdown <input-path-or-url> [output.md]`
+`/to-markdown <input-path-or-url> [output]`
 
 ## Execution
 
 1. Parse `$ARGUMENTS` as:
    - first token: input path or URL
-   - optional second token: output file path
+   - optional second token: output path
+2. If input is a directory:
+   - do not call `Read File` on the directory path
+   - require output to be a directory path and convert supported files recursively
 2. Run:
 
 ```bash
 bash /home/vscode/.opencode/skills/markitdown-converter/scripts/convert.sh <input> [output]
 ```
 
-3. If output path is provided, report the saved path.
+3. If output path is provided, report the saved path(s).
 4. If output path is omitted, return the generated Markdown in the response.

.devcontainer/opencode-seed/instructions/markitdown-converter.md

@@ -2,5 +2,6 @@ When users ask to read, summarize, or analyze non-Markdown sources (PDF, DOCX, P
 
 Guidelines:
 - Prefer `/to-markdown <input> <output.md>` for large inputs.
+- If `<input>` is a directory, never use `Read File` on that directory path; run `/to-markdown <input-dir> <output-dir>` instead.
 - For small inputs, `/to-markdown <input>` and return inline Markdown is acceptable.
 - After conversion, continue analysis using the Markdown output only.

.devcontainer/opencode-seed/skills/markitdown-converter/SKILL.md

@@ -25,3 +25,4 @@ Use this skill when you need to transform non-Markdown sources (PDF, Office docs
 
 - If output is large, prefer writing to a file and then summarizing key sections.
 - Preserve the generated Markdown as an artifact when reproducibility is needed.
+- When the input is a directory, pass an output directory and convert files in bulk instead of reading the directory as a file.

.devcontainer/opencode-seed/skills/markitdown-converter/scripts/convert.sh

@@ -10,11 +10,49 @@ INPUT="$1"
 OUTPUT="${2:-}"
 
 if ! command -v markitdown >/dev/null 2>&1; then
-  echo "markitdown is not installed. Install with: uv tool install markitdown" >&2
+  echo "markitdown is not installed. Install with: uv tool install 'markitdown[all]'" >&2
   exit 2
 fi
 
-if [ -n "$OUTPUT" ]; then
+is_supported_file() {
+  local path="$1"
+  case "${path,,}" in
+    *.pdf|*.doc|*.docx|*.ppt|*.pptx|*.xls|*.xlsx|*.html|*.htm|*.txt|*.csv|*.json|*.xml)
+      return 0
+      ;;
+    *)
+      return 1
+      ;;
+  esac
+}
+
+if [ -d "$INPUT" ]; then
+  if [ -z "$OUTPUT" ]; then
+    echo "Input is a directory. Provide an output directory as the second argument." >&2
+    exit 3
+  fi
+
+  mkdir -p "$OUTPUT"
+  converted=0
+
+  while IFS= read -r -d '' source_file; do
+    if ! is_supported_file "$source_file"; then
+      continue
+    fi
+
+    rel_path="${source_file#$INPUT/}"
+    target_path="$OUTPUT/${rel_path%.*}.md"
+    mkdir -p "$(dirname "$target_path")"
+    markitdown "$source_file" > "$target_path"
+    converted=$((converted + 1))
+    echo "Saved Markdown: $target_path"
+  done < <(find "$INPUT" -type f -print0 | sort -z)
+
+  if [ "$converted" -eq 0 ]; then
+    echo "No supported files found in: $INPUT" >&2
+    exit 4
+  fi
+elif [ -n "$OUTPUT" ]; then
   markitdown "$INPUT" > "$OUTPUT"
   echo "Saved Markdown: $OUTPUT"
 else