/structure-voice-memo

/therapy-tracking:structure-voice-memo

Take an unstructured transcript (or freeform text) and produce a tidy, readable summary the user can hand to their therapist. Organisation only — no interpretation.

Procedure

1. Resolve config: Read $PLUGIN_DATA_DIR/config.json. If missing, tell the user to run onboard first and stop.

2. Source the transcript:

   • If $ARGUMENTS is a file path, read it.
   • Otherwise ask the user to paste the transcript or give a path.

3. Edit for readability (do not change meaning):

   • Remove filler words, false starts, immediate repetitions.
   • Insert paragraph breaks where the user pauses or shifts topic.
   • Fix obvious transcription errors (homophones, mis-segmented words) using context.
   • Keep the user's voice — do not rewrite into clinical or formal register.

4. Extract discrete stressors / topics: List each distinct issue the user named, in their own words. One short bullet each. Do not merge or interpret.

5. Categorise (only if the user asks): group bullets by topic the user themselves named. Never apply diagnostic categories.

6. Write output: Save to <workspace_path>/therapy-sessions/YYMMDD-voice-memo-<slug>.md with two sections:

   • Cleaned transcript — the edited version.
   • Stressors / topics — the bullet list.

7. Report: Print the file path. Suggest /therapy-tracking:log-session pre if the user wants to also write a forward-looking pre-session note.

Rules

• Never interpret feelings for the user. Do not say "this sounds like anxiety" or "you seem to be describing X". Surface only what the user said.
• Never add anything not in the source. No inferred motivations, no synthesised conclusions.
• Preserve voice. The output should still sound like the user, just tidied.