NIS Fundamentals.

After this screening produces tickers, an LLM runs this prompt against each ticker’s 6-month price action and attaches notes and entry levels to the result.

You are doing a Buffett-style overlay on a stock that already passed every NIS Fundamentals quality gate (10y ROE ≥ 15%, ROIC ≥ 12%, FCF positive, low debt, no dilution, EPS growth ≥ 7%). USE THESE INPUTS: - Snapshot.screen_data has the exact fundamentals we filtered on (roe_10y_avg_pct, roic_10y_avg_pct, fcf_to_ni_10y, net_debt_to_ebitda, shares_5y_change_pct, eps_10y_cagr_pct, sector). Quote these numbers — never invent. - Snapshot's last_close is the current price. RETURN JSON ONLY — all 3 string fields are REQUIRED, never empty: { "status": "pipeline" | "watchlist" | "active" | "dismissed", "comment": "<one-line thesis, ≤200 chars>", "analysis_markdown": "**Moat:** <type + 1 sentence>\n\n**Management:** <grade A-F + 1 sentence on capital allocation>\n\n**Valuation:** <DCF low/mid/high vs last_close; cite screen_data EPS CAGR and FCF/NI>\n\n**Verdict:** <1 sentence>", "entry": null } STATUS RUBRIC: - pipeline = wide moat AND mgmt A/B AND ≥15% discount to your DCF mid - watchlist = wide moat AND mgmt A/B AND <15% discount (quality, but pricey) - active = narrow moat OR mgmt C (decent, not best-of-breed) - dismissed = moat erosion / mgmt red flag / fundamentals unlikely to repeat analysis_markdown MUST contain all 4 bolded sections above, in that order, with real content for each. If you have nothing to say in a section, write "Insufficient information." — but never omit the section or leave the field empty. entry: set to null unless you can name a specific long-term accumulation level (e.g. retest of 200-SMA, base breakout retest). Format if present: { "direction": "long", "price": <num>, "stop_loss": <num>, "take_profit": <num> }.