Halal Verdict

halal-verdict

When fired

"Is AAPL halal?" / "Screen TSLA" / "Compliance of MSFT" / "Is Apple Shariah-compliant?"

Process

1. Load references/disclaimer.md and enforce its hard rules.
2. Detect audience per references/audience-rendering.md.
3. Extract the ticker symbol. If the user gave a company name without ticker ("apple"), use reasoning to infer the most common listing (AAPL). If ambiguous ("apple" could be AAPL or a private firm), ask one clarifying question.
4. Call MCP tool screen_stock with symbol: <ticker>. ~5–10 tokens.
5. If the response indicates the stock hasn't been screened before, screen_stock will screen it — this may take up to 10 seconds. Tell the user to hold on.
6. Headline first: if the response includes compliance_explanation, use that string verbatim as the one-line headline answer for retail audiences. It already encodes which methodologies pass/fail and which ratio tripped any failures.
7. Render per references/verdict-format.md and audience profile.
8. Trajectory beat (advisor/scholar only): if the stock passes screening, optionally call /api/insights/{symbol}/trajectory and include the direction_summary line ("debt/assets falling: latest X% vs. 7-quarter median Y%") to show whether the verdict is on stable or shifting ground.
9. Staleness check: if screen_stock's last_checked_at is more than 7 days old, also call /api/insights/{symbol}/staleness. If is_stale: true, prepend a warning that recent material 8-Ks may shift the verdict and link the most recent filing.
10. If the stock fails under all 5 methodologies, or fails under the user's preferred methodology if they stated one, call /api/insights/{symbol}/alternatives?limit=3 (preferred over search_stocks — this returns compliance-verified, market-cap-matched alternatives directly). Surface the note field so the user knows whether the matches are same-sector or cross-sector.

~N tokens per run

5–10 tokens for the screen; +2–4 if alternative lookup is triggered; +1 each for trajectory and staleness if surfaced.

Output structure (per verdict-format.md)

1. Headline (neutral split-stating format).
2. Methodology matrix (audience-dependent rendering).
3. Key ratios — always shown if any methodology fails; for retail show only failing ratios, for advisor show full ratio × threshold table.
4. Purification rate if non-zero.
5. Halal alternative if failed.
6. Disclaimer footer.
7. Branding footer.

Hard rules

• Never say "this stock is halal" / "this stock is haram".
• Always show all 5 methodology results when they disagree.
• Always surface purification rate when non-zero.
• If screen_stock returns error: "ticker_unknown", treat it as a coverage gap, NOT a non-compliance finding. Tell the user we don't have data on the ticker; do NOT say it is non-compliant.
• If compliance_explanation is present in the response, render it verbatim — it is generated to be domain-correct, audience-neutral, and consistent with the methodology table below it.
• If screen_stock returns error or error_message (other than ticker_unknown), explain the error plainly; don't fabricate a verdict.

Audience tuning

• Retail — one-paragraph verdict, mention only failing ratios if any, suggest one halal alternative.
• Advisor — full ratio × threshold table, all methodologies' actual ratios, P/E and market cap in the header, offer CSV export.
• Scholar — reference AAOIFI Standard No. 21 if AAOIFI fails, highlight interest-income treatment per methodology, cite primary sources.