Knowledge Extraction
Apply the source contract
Read ../../../.github/specs/knowledge-extract.spec.md when present. If acquisition or normalization is in scope, also read ../../../.github/specs/source-capture.spec.md. Resolve source scope, mode, depth, organization, adaptation, output profile, privacy boundary, and destination.
Process sources
- Register each source with stable identity, medium, provenance, and availability.
- Normalize format while preserving chronology, authorship, branches, attachments, code, tables, and other meaningful structure.
- Extract bounded knowledge units with type, confidence, time scope, source trail, and contextual qualifiers.
- Separate observed content from inference, recommendation, and unresolved interpretation.
- Reconcile duplicates, contradictions, evolution, supersession, and dependencies.
- Synthesize the requested artifact without severing claims from evidence.
Protect epistemic integrity
- Never invent missing source content.
- Preserve disagreement and uncertainty instead of forcing consensus.
- Treat transient statements, brainstorming, and emotional expression as contextual unless later evidence stabilizes them.
- Distinguish current facts from historical facts, preferences, plans, and abandoned ideas.
- Minimize sensitive personal details to what the requested knowledge artifact needs.
- Keep extraction separate from downstream synapse creation unless both stages are requested.
Validate and deliver
Check source coverage, attribution, chronology, confidence, contradiction handling, privacy, and gaps. Produce the specification's standard output or requested profile, including a source register, context capsule, knowledge inventory, relationships and evolution, decisions and open loops, reusable artifacts, provenance, uncertainty, and omissions.
1---2name: knowledge-extraction3description: Extract durable, attributable, reusable knowledge from conversations, documents, code, repositories, web pages, transcripts, images, audio, video, or mixed source sets. Use for knowledge inventories, source-grounded summaries, decision extraction, chronology, reusable artifacts, contradictions, open loops, or Mind Garden preparation.4---56# Knowledge Extraction78## Apply the source contract910Read `../../../.github/specs/knowledge-extract.spec.md` when present. If acquisition or normalization is in scope, also read `../../../.github/specs/source-capture.spec.md`. Resolve source scope, mode, depth, organization, adaptation, output profile, privacy boundary, and destination.1112## Process sources13141. Register each source with stable identity, medium, provenance, and availability.151. Normalize format while preserving chronology, authorship, branches, attachments, code, tables, and other meaningful structure.161. Extract bounded knowledge units with type, confidence, time scope, source trail, and contextual qualifiers.171. Separate observed content from inference, recommendation, and unresolved interpretation.181. Reconcile duplicates, contradictions, evolution, supersession, and dependencies.191. Synthesize the requested artifact without severing claims from evidence.2021## Protect epistemic integrity2223- Never invent missing source content.24- Preserve disagreement and uncertainty instead of forcing consensus.25- Treat transient statements, brainstorming, and emotional expression as contextual unless later evidence stabilizes them.26- Distinguish current facts from historical facts, preferences, plans, and abandoned ideas.27- Minimize sensitive personal details to what the requested knowledge artifact needs.28- Keep extraction separate from downstream synapse creation unless both stages are requested.2930## Validate and deliver3132Check source coverage, attribution, chronology, confidence, contradiction handling, privacy, and gaps. Produce the specification's standard output or requested profile, including a source register, context capsule, knowledge inventory, relationships and evolution, decisions and open loops, reusable artifacts, provenance, uncertainty, and omissions.
egohygiene