Do equity research with Claude Code without coding
A non-engineer keeps research in a folder of plain-text files managed with Obsidian and uses Claude Code for equity research. Task: Suggested implementation: A non-engineer organizes SEC filings and research notes in a local Obsidian folder with a CLAUDE.md file defining structure and rules. Claude Code reads these files to extract financial figures, attaching exact quoted lines and accession numbers for verification. The human role involves setting up the folder structure, defining the CLAUDE.md rules, and reviewing the extracted data for accuracy.; tools: Claude Code. Human role: The original post does not say how a person stayed involved.; limitations: Prerequisites include a local folder structure, Obsidian, and SEC filings from EDGAR. Constraints involve strict workflow discipline: calculations must be scripted with quoted inputs, and long filings may require chunking. Source reliability is low due to a 403 error; human approval is required for all data extractions before use in analysis..
Tools
Published on Agent Work Atlas:
Overview
A non-engineer keeps research in a folder of plain-text files managed with Obsidian and uses Claude Code for equity research.
Workflow
Suggested implementation: A non-engineer organizes SEC filings and research notes in a local Obsidian folder with a CLAUDE.md file defining structure and rules. Claude Code reads these files to extract financial figures, attaching exact quoted lines and accession numbers for verification. The human role involves setting up the folder structure, defining the CLAUDE.md rules, and reviewing the extracted data for accuracy.
Reported result
Actual outcomes are unknown as the source is unavailable. Proposed measurable acceptance check: Verify that 100% of extracted figures match the original text and include valid accession numbers. A labeled test case would be to identify a specific metric trend across five earnings releases, such as a bookings metric changing from $3B to $5.9B before disappearing.
Human role
The original post does not say how a person stayed involved.
Limitations
Prerequisites include a local folder structure, Obsidian, and SEC filings from EDGAR. Constraints involve strict workflow discipline: calculations must be scripted with quoted inputs, and long filings may require chunking. Source reliability is low due to a 403 error; human approval is required for all data extractions before use in analysis.
Try it
Act as an equity research assistant. I have a local folder [replaceable: path to Obsidian vault] containing SEC filings [replaceable: list of accession numbers or file names] and a CLAUDE.md file [replaceable: path to CLAUDE.md]. First, read the CLAUDE.md to understand the file structure and extraction rules. Then, process the specified filings to extract key financial metrics. For each metric, provide the value, the exact quoted line from the source, and the accession number. Do not perform calculations; only extract raw data. Output format: A markdown table with columns [Metric, Value, Source Quote, Accession Number]. Before processing all files, run a dry test on one filing [replaceable: specific file name] and show me the output for confirmation. If the output is correct, proceed with the rest. If inputs are missing, ask for clarification.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.