clotho-daily-debrief

name	clotho-daily-debrief
description	Use when the user says 'end of day', 'daily debrief', 'process today', 'what happened today', 'dump my day', 'debrief', or wants to capture everything from today into Clotho.

Daily Debrief

End-of-day ceremony. Your job is to get the user's entire day captured, structured, and linked in Clotho. You handle four phases: Intake, Status Update, Horizon Check, then launch the debrief-processor agent for extraction.

The workspace is set automatically. Use clotho_set_workspace if needed.

Phase 1: Intake

Step 1: Check what's already here

1. Scan .clotho/inbox/ for unprocessed files:

   Use Bash to: ls -la <workspace_path>/.clotho/inbox/
   

   If files exist, report what's there.

2. Query entities created today:

   clotho_list_entities(entity_type: "Meeting")
   clotho_list_entities(entity_type: "Transcript")
   clotho_list_entities(entity_type: "Note")
   

   Filter results to items with today's date in created_at.

3. Present to user:

   "Here's what I already have from today: [list meetings/transcripts/notes]. Your inbox has N unprocessed files."

Step 2: Capture inbox

For each file in .clotho/inbox/:

• Infer type from filename/extension (.transcript.md → transcript, .md → note, .ics → meeting)
• Capture: clotho_capture(file_path, entity_type, title)
• After capturing, move the file out of inbox (or note it's been processed)

Step 3: Gather additional materials

Ask the user:

"Were there any other meetings or conversations today not captured above? You can:

• Paste a transcript or meeting summary
• Point me to a file
• Just tell me what happened and I'll capture it as a note"

For each material provided:

• Pasted text → clotho_create_note(title, content) or clotho_create_entity(entity_type: "transcript", title, content)
• File reference → clotho_capture(file_path, entity_type)
• Verbal description → clotho_create_note(title: "EOD note: <topic>", content: <what they said>)

Keep asking until the user says they're done.

Phase 2: Status Update

Step 4: Task review

1. Query active and pending tasks:

   clotho_list_entities(state: "doing")
   clotho_list_entities(state: "todo")
   

2. Present as a bulk list:

   "Here are your active/pending tasks. What moved today?"

   Doing:

   • [title] (id)
   • [title] (id)

   Todo:

   • [title] (id)

   Tell me what changed: completed, blocked, started, or no change.

3. For each update the user provides:

   • Completed → clotho_update_entity(entity_id, state: "done")
   • Blocked → clotho_update_entity(entity_id, state: "blocked") + ask what's blocking, create Blocker entity + BLOCKED_BY relation
   • Started → clotho_update_entity(entity_id, state: "doing")

Step 5: Ad-hoc items

Ask:

"Any decisions, risks, or blockers from outside meetings today?"

For each:

• Decision → clotho_create_entity(entity_type: "decision", title)
• Risk → clotho_create_entity(entity_type: "risk", title)
• Blocker → clotho_create_entity(entity_type: "blocker", title)

Link to relevant programs if the user specifies context.

Phase 3: Horizon Check

Step 6: Look ahead

1. Query tasks with upcoming deadlines:

   clotho_query(cypher: "MATCH (t:Task) RETURN t.id, t.title, t.entity_type")
   

   Check metadata for deadline fields. Also:

   clotho_list_entities(state: "todo")
   

   Check created_at — flag any todo tasks older than 7 days as stale.

2. Check program health:

   clotho_list_entities(entity_type: "Program")
   

   For each program, check if it has any tasks updated in the last 2 weeks. Flag programs with no recent activity.

3. Present:

   "Looking ahead:

   • N items due this week
   • M items due next week
   • K tasks have been in todo for >7 days: [list]
   • Program X has had no task activity in 2 weeks

   Want to reprioritize anything or flag something?"

Handle any reprioritization the user requests.

Phase 4: Extract & Wrap

Step 7: Launch agent

Launch the debrief-processor agent. Pass it the context of which transcripts/notes from today haven't been extracted yet (no EXTRACTED_FROM relations pointing to them).

The agent will:

• Read transcript/note content
• Identify speech acts (decisions, risks, tasks, blockers, questions, insights)
• Create derived entities with EXTRACTED_FROM relations
• Identify and link people (MENTIONS)
• Suggest BELONGS_TO relations to programs

Step 8: Review and sync

After the agent completes:

1. Present the extraction summary
2. Ask: "Does this look right? Anything to adjust?"
3. Handle any corrections
4. Sync: clotho_sync()
5. Confirm: "Today is captured. Good night."