Rapid Context Extractor

Rapid Context Extractor

Normalize source mechanics with the script, then do semantic analysis in chat.

Quick Start

python3 Skills/rapid-context-extractor/scripts/prepare_payload.py \
  --seed-file "./Research/topic-frame.md" \
  --source-url "https://example.com/article" \
  --auto-semantic \
  --output "/home/.z/workspaces/<conversation-id>/extraction_packet.md"

Replace <conversation-id> with your active conversation workspace, or use any other writable output path.

Use one source input per run:

• --source-url for web pages
• --source-file for local docs/transcripts/subtitles
• --source-text for pasted text

Optional seed context:

• --seed-file or --seed-text

Optional semantic memory anchoring:

• --semantic-query to retrieve relevant prior concepts from your semantic memory
• --auto-semantic to generate semantic query from source title + extracted terms (recommended default)
• --semantic-limit (default 5) to control number of memory anchors
• --provenance to force frontmatter provenance (otherwise inferred from output path conversation ID)

Workflow

1. Prepare packet

• Run scripts/prepare_payload.py to produce a markdown packet containing seed context + chronological source chunks.
• For media files, require transcript sidecar or transcribe first.

2. Adopt analyst frame

• Read seed context first.
• State the frame in 1-2 lines before distillation.
• If missing background blocks understanding, perform targeted research before summarizing.

3. Distill in chronological order

• Produce bullet points in the order ideas appear in source.
• Avoid regrouping by theme if it breaks chronology.
• Keep claims faithful to the source.

4. Include image meaning

• If visuals exist, summarize what each visual contributes to the argument.
• Note if visuals reinforce, contradict, or extend text claims.

5. Integrate for learning

• Explain key terms, concepts, and implications in plain language.
• Connect key claims to Semantic Memory Anchors where relevant (agreements, tensions, extensions).
• Explicitly classify each integration claim as aligns, extends, or conflicts/tension.
• Ask clarifying questions that advance interpretation or decisions.

6. Force active engagement

• Ask for immediate reaction (1-3 lines acceptable).
• Ask for one agreement and one challenge.
• Offer optional ingestion: only ingest if user explicitly says yes.

Standard Output Shape

Use this structure in responses:

1. Analytical Frame
2. Chronological Distillation
3. Visual Layer (if applicable)
4. Semantic Integration (link to user- or project-specific anchors when available) : include explicit aligns / extends / conflicts labels
5. Concept Decoder
6. Clarifying Questions
7. Your Reaction (collect user response)
8. Optional Next Step (ingest yes/no)

Content Library Ingestion

Only after explicit approval, and only in workspaces that include the N5 ingestion helper:

python3 N5/scripts/content_ingest.py "<artifact_path>" --move

Confirm with: Ingested to Content Library as <type>.

Resources

• scripts/prepare_payload.py: deterministic intake/normalization for seed + source.
• references/output-template.md: copyable response template for consistent execution.