Public Records Researcher

Step 2: Check Dependencies

python3 "$SKILL_DIR/scripts/check_dependencies.py"

• Exit 0: all good. Exit 1: packages were installed (proceed). Exit 2: failed (report to user).

Step 3: Run Records Researcher

WORK_DIR="$(pwd)/legal-records-$(date +%s)"
mkdir -p "$WORK_DIR"

python3 "$SKILL_DIR/scripts/research_records.py" \
  --company "<company_name>" \
  --output-dir "$WORK_DIR" \
  [--cik <number>] \
  [--filing-types 10-K,10-Q] \
  [--years 5]

The script outputs JSON to stdout with the research results.

Step 4: Present Results

1. Read $WORK_DIR/research_summary.txt and present the research findings to the user.
2. Highlight key findings:
   • Company identification (name, CIK, SIC code, state of incorporation)
   • Number and types of filings found
   • Key officers and directors
   • Financial highlights (revenue, net income trends)
   • Notable risk factors
3. Inform the user that the following files were generated:
   • company_profile.json - structured company data
   • filings_list.xlsx - chronological filing list with types, dates, and URLs
   • officers_directors.xlsx - officer and director information
   • financial_trends.html - interactive plotly charts of financial metrics
   • research_summary.txt - human-readable research report

Step 5: Offer Additional Actions

• View financial charts: Point user to $WORK_DIR/financial_trends.html to open in browser
• DOCX report: Offer to generate a formal research report (.docx) using the npm docx package
• Deep dive: If user wants details on a specific filing, offer to retrieve and analyze it
• Refine search: If user wants different filing types or time period, loop back to Step 3

Accuracy and QA (Required)

Anti-hallucination rules (include in ALL subagent prompts):

• Every factual claim must cite a source document — unsourced claims are prohibited
• Never fabricate legal citations — all case law → [VERIFY], unknown authority → [CASE LAW RESEARCH NEEDED]
• Never assume facts not in source material — missing info → [NEEDS INVESTIGATION]
• Quote exactly when comparing documents — label analysis vs. facts distinctly

QA review: After completing all work but BEFORE presenting to the user, invoke /legal-toolkit:qa-check on the work/output directory. Do not skip this step.

Error Handling

• Company not found: Suggest checking the spelling or using the CIK number directly
• No filings found: The company may not have SEC filings (private company); inform the user
• Network error: SEC EDGAR requires internet access; ask user to check connectivity
• Rate limiting: SEC EDGAR has rate limits; the script handles retries automatically
• Script not found: Verify the skill is installed (ls $SKILL_DIR/scripts/)