Back to catalog / Agent workflows

Distill SEC filings into a provenance wiki

Fetch a set of public SEC filings once, distill them into traceable Markdown wiki pages, and answer later questions with fact provenance.

Goal

Build a reusable source-linked wiki for recurring questions and expose conflicts across documents.

Choose a tier with build_manifest.py, then fetch public SEC documents by accession using fetch_data_room.py to create provenance-stamped text. [src10-anthropic-distill-sec-filings-into-provenance-wiki]
Extract entities, timeline, amounts, and relationships by section into Markdown wiki pages in the Managed Agents memory store; retain document/page citations. [src10-anthropic-distill-sec-filings-into-provenance-wiki]
Query the wiki with cases for known price changes, look-alike entities, and a synthetic conflicting file; check each fact links to its SEC source. [src10-anthropic-distill-sec-filings-into-provenance-wiki]

Key amounts/timeline facts resolve to SEC accession and page; synthetic conflicts are flagged rather than overwriting filings.

For missing citations, inspect manifest/page retention; if a conflict is collapsed, separate document timelines and preserve both sources.

Claude Managed Agents access, public SEC filings, and model permissions; tier resources and access depend on current preview availability.

Reads public SEC filings and sends content to Anthropic models; writes wiki pages to Managed Agents memory and incurs storage/inference costs.

The example deal materials are illustrative, not financial/legal/investment advice; facts absent from filings may remain unknown.

These values describe the upstream repository, not popularity or likes for an individual skill or playbook.

No metrics snapshot is available yet.

License: MIT