nac
777607dc29
Track the source-of-truth code, skills, bootstrap rules, docs, and packaging files before adding new features. Exclude local vault config, generated skill mirrors, temporary exports, virtual environments, and business data files from version control. Validation: inspected staged file list and git ignored status before commit.
1.8 KiB
Ingest Prompt Templates
These are the mental frameworks to use when distilling a source into wiki pages.
Knowledge Extraction Frame
When reading a source document, ask yourself:
-
What are the 3-5 most important ideas in this document? These become concepts pages or updates to existing concept pages.
-
Who or what is mentioned that deserves its own page? People, tools, organizations, projects → entity pages.
-
What does this document teach you how to do? Procedures, workflows, techniques → skills pages.
-
What claims does this document make? Each claim needs a source attribution. If it contradicts an existing wiki claim, note the contradiction.
-
How does this connect to what the wiki already knows? This is the most important question. The value of the wiki compounds through connections.
Synthesis Frame
When a new source covers ground that existing pages already cover:
- Don't duplicate — synthesize
- If the new source agrees with existing content, strengthen the claims with additional attribution
- If it disagrees, create an "Open Questions" or "Debate" section noting both positions
- If it adds nuance, weave it into the existing narrative
Cross-Reference Discovery
After extracting knowledge, look for these connection patterns:
- Is-a: "Transformers are a type of neural network" → link from transformer page to neural-network page
- Uses: "RLHF uses reward models" → link from RLHF to reward-models
- Contrasts-with: "CNNs vs. Transformers for vision" → mutual links
- Part-of: "Attention is a component of transformers" → link from attention to transformers
- Created-by: "Transformers were introduced by Vaswani et al." → link to entity page
- Applied-in: "Transformers are used in GPT" → link from transformers to GPT