Menu
Use when the user says 'process this transcript', 'extract from this meeting', 'capture transcript', 'here's a transcript', or provides a meeting transcript/notes for extraction. Also triggered when a file is provided with meeting content.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Use when the user asks to 'init clotho', 'create a workspace', 'capture a file', 'create a program', 'create a task', 'add a person', 'list entities', 'delete entity', 'update entity', or any entity CRUD operation in a Clotho workspace.
Use when the user says 'weekly review', 'weekly reflection', 'review my week', 'end of week', 'retrospective', or wants to reflect on the past week's work.
Use when the user asks to 'extract from transcript', 'find decisions', 'identify risks', 'what action items', 'extract insights', 'process meeting notes', or wants AI-assisted extraction of structured entities from unstructured content.
Use when the user says 'morning', 'daily brief', 'what should I focus on', 'start of day', 'what's on today', or wants a prioritized view of today's work.
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.
Use when the user asks to 'relate entities', 'create a relation', 'show relations', 'query the graph', 'what belongs to', 'what's blocking', 'who is mentioned', or any graph/relation operation in Clotho.
| name | clotho-transcript-processor |
| description | Use when the user says 'process this transcript', 'extract from this meeting', 'capture transcript', 'here's a transcript', or provides a meeting transcript/notes for extraction. Also triggered when a file is provided with meeting content. |
Single-transcript processing. The user has a transcript or meeting notes — process it, extract signals, and link everything. This is the manual/ad-hoc version of what the daily debrief does in batch.
The workspace is set automatically. Use clotho_set_workspace if needed.
Determine how the user is providing the content:
File path:
clotho_capture(file_path: "<path>", entity_type: "transcript", title: "<meeting name>")
Pasted text:
clotho_create_entity(entity_type: "transcript", title: "<meeting name>", content: "<pasted text>")
Verbal description:
clotho_create_note(title: "<meeting name> - Notes", content: "<what they described>")
If the user doesn't provide a title, infer from the content (first line, meeting subject, attendee names + date).
Ask (or infer from content):
"Which program does this meeting relate to?"
Present active programs:
clotho_list_entities(entity_type: "Program", status: "active")
If the user specifies a program, read its content to understand the extraction lens:
clotho_read_entity(entity_id: "<program_id>")
If the user says "not sure" or it spans multiple, proceed with general extraction and route later.
If this was a meeting (not just standalone notes), ensure a Meeting entity exists:
clotho_create_entity(entity_type: "meeting", title: "<meeting name>")
Link the transcript to the meeting:
clotho_create_relation(source_id: "<transcript_id>", relation_type: "spawned_from", target_id: "<meeting_id>")
Launch the debrief-processor agent with:
The agent will:
Present the agent's extraction summary. Ask:
"Does this capture everything? Anything to add or adjust?"
Handle corrections, then:
clotho_sync()
If the user asks to process a transcript that's already been extracted (has EXTRACTED_FROM relations pointing to it), warn:
"This transcript has already been processed. N entities were extracted from it. Want to re-extract (may create duplicates) or skip?"
Check with:
clotho_get_relations(entity_id: "<transcript_id>")
Look for incoming EXTRACTED_FROM relations.