Agent skill · writing content · onewave-ai
meeting-to-tasks
Takes meeting transcripts, extracts action items with owners and deadlines, detects implicit commitments, generates structured meeting summaries, and outputs task files compatible with Linear, GitHub Issues, and other project management tools.
Why this skill is useful
Adds a comprehensive workflow for transforming meeting transcripts into structured task outputs, including scripts for task generation and confidence scoring that are not commonly available.
What it needs
Requires bash, grep installed locally. About 2k tokens when loaded. Last updated 2026-07-15. 242 stars on the source repository.
What this skill does
Meeting to Tasks Transform raw meeting transcripts into structured, actionable outputs: decisions, action items, owners, deadlines, and open questions -- including implicit commitments participants may not realize they made. Contents references/extraction-schema.md -- YAML field schemas for every extraction phase references/markers.md -- Decision/action/implicit markers, priority + deadline inference, confidence scoring references/output-templates.md -- Meeting summary, task files (Linear/GitHub/generic), follow-up email, directory layout references/meeting-types.md -- Meeting-type detection and per-type focus Workflow 1. Accept input. Handle plain text, Markdown, pasted text, audio transcript exports (Otter.ai, Fireflies, Rev, Zoom, Google Meet, Microsoft Teams), or structured notes. 2. Validate. Confirm there is enough content (~100 words minimum), identifiable speakers, a discernible topic, and any timestamps. If input is too sparse, ask for the full transcript, the participant list, or the meeting purpose before proceeding. 3. Read the full transcript first. Do not extract from a partial read -- later context can change earlier interpretation. 4. Detect the meeting type and adjust focus per references/meeting-types.md. 5. Extract in phases using the schemas in references/extraction-schema.md: metadata, decisions, action items, open questions, parking lot, discussion points. Apply the markers and inference rules in references/markers.md. Preserve the exact source quote for every item. 6. Identify speakers. Build the participant list from speaker labels and track who said what. When labels are absent, infer from context clues, otherwise mark "Unassigned" and flag for the user to assign. 7. Score confidence on every item (high/medium/low). Flag all low-confidence items for human review per references/markers.md. 8. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills onewave-ai/meeting-to-tasks