the task takes a destination

output_dir, optional. absent, the text lands beside the recording as it
always has — that is what a person right-clicking a memo wants and it is
not changing. given, the transcript goes there instead, and the directory
is created if it does not exist.

for callers that are not a person clicking a file. a commander keeps its
results in its own directory, and the alternative was making every
commander move the file afterwards: a step in the loop forever, plus a
window where the file sits in the folder being watched.

marked `hidden` so the file browser neither shows it nor counts it as
something to ask about — without that, declaring it turned a direct
right-click run back into a modal with a field nobody clicking a file has
an answer for.

tasks/transcribe-audio-file/TASK.md

@@ -15,6 +15,13 @@ inputs:
   - name: file_path
     description: Path to the audio file to transcribe. Auto-filled from the file browser.
     required: true
+  - name: output_dir
+    description: >-
+      Where to write the transcript. Absolute, or relative to the recording. Leave it out and the text
+      lands beside the recording, which is what a person clicking a file wants. A commander passes its own
+      directory so its results never land in the folder it watches.
+    required: false
+    hidden: true
 trigger:
   - type: file
     extensions:
@@ -38,8 +45,8 @@ model call spent re-deriving what the plugin already returns.
    (`GET /api/user/settings`).
 2. `POST` the file to `/api/transcribe/transcribe`. The plugin detects the language and transcribes
    faithfully in it — see `plugins/transcribe/skills/transcribe/SKILL.md` for the contract.
-3. Write the transcription beside the audio file, same name, `.md`. Nothing is moved and nothing is
-   created around it.
+3. Write the transcription as `.md` with the same base name — beside the recording by default, or into
+   `output_dir` when one is given. The directory is created if it does not exist.
 
 It overwrites a previous transcription of the same file, deliberately: a transcription is derived, and
 running it again means you wanted it again.

tasks/transcribe-audio-file/index.ts

@@ -8,7 +8,8 @@
 // only owns filing the result.
 
 import { existsSync } from 'node:fs';
-import { basename, dirname, join, extname } from 'node:path';
+import { mkdir } from 'node:fs/promises';
+import { basename, dirname, isAbsolute, join, extname } from 'node:path';
 
 const filePath = process.env.INPUT_FILE_PATH;
 if (!filePath) {
@@ -69,14 +70,31 @@ async function main() {
   const { text, detectedLanguage } = await transcribe(path, spokenLanguages);
   console.log(`Detected language: ${detectedLanguage}`);
 
-  // Beside the audio, same name, `.md`. Nothing is moved, nothing is created around it — this is the
-  // behaviour the file viewer's transcribe button had, and the one thing anybody wants from a
-  // transcription is the text next to the recording it came from.
-  //
-  // It overwrites a previous transcription of the same file deliberately: a transcription is derived,
-  // running it again means you wanted it again, and a `-2` beside it would only be a second file to
-  // read and choose between.
-  const transcriptPath = join(dirname(path), `${basename(path, extname(path))}.md`);
+  /*
+    Where the text goes.
+
+    By default, beside the audio with the same name — the behaviour the file viewer's transcribe button
+    had, and what a person right-clicking a recording wants: the text turns up next to the thing it came
+    from.
+
+    `output_dir` overrides it, and exists for callers that are not a person clicking a file. A commander
+    keeps its results in its OWN directory: writing them into the folder it watches would put its work in
+    the middle of its subject, where nobody asked for it and where its next pass has to tell the two
+    apart. Passing a destination is one input; making every such caller move the file afterwards is a
+    step in every loop, forever, and a window where the file is in the wrong place.
+
+    Created if missing, so a caller does not have to make the directory first — the first run of a new
+    commander otherwise fails on a directory only this script knows the name of.
+
+    Either way it overwrites a previous transcription of the same recording, deliberately: a
+    transcription is derived, running it again means you wanted it again, and a `-2` beside it would only
+    be a second file to read and choose between.
+  */
+  const outputDir = process.env.INPUT_OUTPUT_DIR?.trim();
+  const destDir = outputDir ? (isAbsolute(outputDir) ? outputDir : join(dirname(path), outputDir)) : dirname(path);
+  if (outputDir) await mkdir(destDir, { recursive: true });
+
+  const transcriptPath = join(destDir, `${basename(path, extname(path))}.md`);
   await Bun.write(transcriptPath, text);
 
   console.log(transcriptPath);