+7

muse/entities.md

reviewed

··· 30 30 Skip entities that don't fit one of these four types. 31 31 Skip any url, domain, filename, or path. 32 32 33 33 + ## Name Resolution 34 34 + 35 35 + - Prefer full names over nicknames or abbreviations when context makes identity clear (e.g., "Dens" → "Dennis Crowley" if the surrounding conversation confirms identity). 36 36 + - Consolidate name variants within the same transcript to a single canonical entity using the most complete name (e.g., "JB", "John B", "John Borthwick" → extract once as "John Borthwick"). 37 37 + - Only extract first-name-only references when identity is unambiguous (one "Sarah" in the conversation). Skip ambiguous first-name references rather than guessing. 38 38 + - Normalize company name variants to canonical form (e.g., "MS", "Microsoft", "MSFT" → "Microsoft"). 39 39 + 33 40 Output as a markdown list. Each line has three parts separated by colon and dash: 34 41 * Type: Entity Name - Description 35 42

+4 -1

muse/knowledge_graph.md

reviewed

··· 34 34 * `First Appearance`: Time. 35 35 * `Total Engagement`: Approximately how many times was it mentioned throughout the day. 36 36 * `Context`: Provide a concise (1-2 sentence) summary of its role or context in how it was used, referencing key actions or discussions from both audio and screen data if distinguishable. 37 37 + * **Concept Quality Filter:** Distinguish `Concept` from `Topic`. A Concept must be a genuinely reusable idea — a mental model, framework, strategic insight, or principle — valuable to recall in a different context. "Discussed the migration timeline" is a Topic (use `Topic` type). "Conway's Law applies to their API design" is a Concept (use `Concept` type). When extracting Concepts, capture WHY the concept matters in the `Context` field, not just that it was mentioned. 37 38 38 39 2. **Relationship Mapping:** 39 40 * Identify and map all significant connections between entities. ··· 59 60 60 61 **Key Considerations:** 61 62 * Synthesize information from all transcript content within each chunk. 62 62 - * Disambiguate entities: e.g., "John" referring to "John Doe." 63 63 + * Disambiguate entities by consolidating name variants to a single canonical entity using the most complete name when the same entity is referenced by different names within the transcript (e.g., "John", "John D.", "John Doe" → use "John Doe" throughout). 64 64 + * When first-name-only references are ambiguous, note the ambiguity in the entity's Context field rather than guessing identity. 65 65 + * Cross-reference names against attendees and participants mentioned earlier in the transcript for spelling corrections and consistent naming. 63 66 * Infer implicit relationships where explicit statements are lacking but context strongly suggests a connection. 64 67 * Focus on the most relevant and significant entities and relationships to avoid an overly noisy graph. 65 68 * For live capture, $preferred often multi-tasks — e.g., joined on a team zoom in the background while working on an unrelated task — so different content streams may not always align.

+2

muse/meetings.md

reviewed

··· 53 53 - Decisions made during the meeting. 54 54 - Action items or follow-ups assigned to specific people. 55 55 - Open questions left unresolved. 56 56 + - Notable ideas or proposals floated, even if not decided on (speculative ideas from meetings are often the most valuable thing to capture — they don't survive as action items but represent new thinking). 56 57 57 58 ## Exclusions 58 59 ··· 69 70 - **Topics Discussed** – concise summary of key subjects and entities mentioned 70 71 - **Slides Presented** – yes/no, with short description if yes 71 72 - **Key Outcomes** – decisions, action items, and open questions 73 73 + - **Key Quotes** – 2-3 memorable or significant quotes with speaker attribution. Omit this section entirely if nothing stands out; do not pad with unremarkable quotes. 72 74 73 75 Conclude with a brief summary (<= 100 words) of the meeting's significance and any immediate next steps. 74 76