The problem

Meeting summaries often suffer from AI hallucinations, vague action items, and a lack of accountability. Developers and managers frequently receive "hallucinated" owners or deadlines that were never actually agreed upon during the discussion.

What it does

• Classifies meetings by risk level, triggering "Human Verification Gates" for high-stakes financial, legal, or security decisions.
• Distinguishes between facts (Decided) and possibilities (Proposed, Open, Rejected) to prevent false assumptions.
• Identifies missing owners and deadlines using "Not specified" instead of guessing, forcing human clarification where gaps exist.
• Generates Mermaid.js workflow diagrams to visualize execution sequences derived from the transcript.
• Produces a structured Operational Intelligence Report including a Master Action Tracker and Risk Mitigation plan.

Frameworks & tools

Compatible with any LLM supporting system instructions (Claude, ChatGPT, OpenClaw). Supports Mermaid.js for visual workflow rendering.

Why this beats prompting it yourself

This skill implements a strict Factual Integrity Protocol that overrides the LLM tendency to be "helpful" by making things up. It uses a dual-category risk system (Category A/B) that ensures sensitive decisions are never treated as routine administrative tasks.

Use cases

• Converting technical syncs into high-integrity Jira or GitHub issue backlogs.
• Auditing security or compliance meetings where "who said what" has legal implications.
• Turning messy brainstorming sessions into clear, prioritized MoSCoW or P0-P3 roadmaps.
• Tracking unresolved dependencies that usually get lost in standard meeting notes.

Known limitations

Requires a raw transcript and organizational context to function. It will halt execution if core variables like priority frameworks are missing.