Source Analyzer

Extracts key claims, quantitative facts, and KQ relevance from user-provided documents/webpages and writes them into source_extracts in YAML.

Overview
Version
v1.0.0
Created
2025-12-14
Updated
2025-12-14
desktop-researchsource-analysisevidence-extractionyamlkq-mapping
source-analyzersa
Key functions
  • Read user-provided documents (text/HTML) and extract key_points
  • Extract quantitative and concrete facts (data_points) as accurately as possible from the original text
  • Map each key/data point to relevant KQs via relevance_to_KQs
  • Write/update only the source_extracts section and output a single updated YAML block
  • Strict formatting in Structured Mode (structured_pipeline: true) with minimal narration
Technical details
_id
g-6916fc70a38c81918caeced7512a7fd3
gpt_id
g-6916fc70a38c81918caeced7512a7fd3
viz1
public
viz2
show_url
language
en
Other fields
additional_features
["KQ-linked structuring (relevance_to_KQs) optimized for handoff to a separate Synthesis GPT"]
example_commands
["Here is my YAML plus a URL/body text—populate only source_extracts and return the updated YAML.", "From this HTML text, extract key_points and data_points and map them to KQ1–KQ3 in source_extracts.", "Extract only numeric stats, percentages, dates, and concrete factual statements into data_points as precisely as possible."]
gpt_id
g-6916fc70a38c81918caeced7512a7fd3
ideal_use_cases
["Structuring key claims and numbers from multiple sources for downstream use", "Tagging/mapping source content to predefined KQs (research questions)", "Preparing standardized source_extracts prior to a separate synthesis step"]
limitations
["Does not discover new sources or modify/add to the sources list", "No synthesis (no comparison, conclusions, or recommendations) — extraction only", "Cannot directly read PDFs/images/downloaded files; requires the text to be pasted/uploaded", "Output is restricted to a single ‘updated YAML block’ (minimal explanation)"]
target_users
["Desktop research / literature review practitioners", "Evidence-driven writers (policy, market, academic)", "Teams automating the ‘source extraction’ step in a research pipeline"]