meeting-action-items-extractor

When to invoke

• When turning raw meeting transcripts into assigned tasks.
• When capturing decisions and follow-ups for project execution.

Inputs needed

• --input: Path to a plain-text transcript.
• Optional --out: Path to write JSON output.

Workflow

1. Split transcript into lines.
2. Detect action items using heuristics (e.g., "TODO", "Action:", "I'll", "We should", "Can you").
3. Attempt owner extraction from speaker prefix patterns like Name:.
4. Detect decisions using patterns ("Decision:", "We decided").
5. Output a normalized task list.

Output format

JSON with:

• decisions: list of strings.
• action_items: list of {owner, task, evidence_line}.