diff --git a/worker_plan/worker_plan_api/prompt/data/simple_plan_prompts.jsonl b/worker_plan/worker_plan_api/prompt/data/simple_plan_prompts.jsonl
index 1b24c3c8..0062bf6c 100644
--- a/worker_plan/worker_plan_api/prompt/data/simple_plan_prompts.jsonl
+++ b/worker_plan/worker_plan_api/prompt/data/simple_plan_prompts.jsonl
@@ -1,3 +1,4 @@
+{"id": "814faadf-5e19-4d1d-89e0-0689ca2bbc4f", "prompt": "Establish \"Crocodile Security\", headquartered in a converted Ottoman-era fortress on an island in the Nile south of Cairo, where the boardroom floor is a single sheet of reinforced glass suspended directly above the demonstration pool and prospective clients negotiate price while twelve adult Nile crocodiles circle beneath their feet. The founder, a reclusive Egyptian excavation magnate, believes perimeter security lost its way when the world abandoned the moat, and intends to correct history. The company designs, excavates, and stocks live-crocodile security moats as a turnkey service for those who understand that a fence merely delays an intruder, but a moat makes a statement. The market was validated in July 2026: Israel's National Security Ministry budgeted NIS 21 million (about 7 million USD) for a 170-meter crocodile moat at Ketziot Prison, with construction underway as of 22 July 2026, inspired by Florida's \"Alligator Alcatraz\" detention facility, and Israel's environment ministry reclassified the Nile crocodile as a \"tended wild animal\" that same month to make it legal. Where one government leads, others will follow, and a specialist vendor beats every interior ministry improvising its own reptile program. The Ketziot benchmark works out to roughly 41,000 USD per meter of stocked moat, and Crocodile Security will win contracts by bidding 20 percent below it. Clientele spans corrections ministries, royal palaces, billionaire island compounds, doomsday bunker communities, gold vaults, and superyacht marinas; the company maintains a strict policy that no client is too controversial, only insufficiently ambitious. The core offering is the turnkey live-crocodile moat: site survey and moat engineering, excavation and water management in arid climates, sourcing Nile crocodiles from Egypt's crocodile farms around Lake Nasser, habitat design with basking zones and thermal management, veterinary care, feeding logistics, handler training in ceremonial black uniforms, escape-prevention engineering, and multi-year maintenance contracts covering animal replacement and water quality, wrapped in an independently audited animal welfare program that doubles as the company's PR shield. A secondary tier offers piranha-stocked moats for narrower channels, smaller budgets, or jurisdictions whose wildlife rules cover large reptiles but not fish, and a saltwater variant is under study for coastal fortresses. A third revenue line is regulatory consulting, helping client governments amend wildlife classifications of the \"tended wild animal\" variety so they can buy the product. Seed budget is 10 million USD, from founder capital plus a Gulf family-office investor, deployed in gated tranches of 2, 3, and 5 million. The first paying contract is due within 18 months, with the Ketziot-style deal (5 to 7 million USD) as the reference size. Success criteria: three signed contracts or funded pilots by month 24, CITES-compliant export permits for Nile crocodiles secured in year one, zero animal welfare violations, zero human injuries, positive operating cash flow by month 30, and a stocked moat on every inhabited continent by year seven. Pick a realistic scenario, not the most aggressive one; this is a first venture.", "tags": ["egypt", "crocodile", "moat", "security", "prison", "animal", "business"]}
 {"id": "0e8bc3e2-48c6-4a33-93c5-73205d5918e9", "prompt": "Establish an intentionally extreme hyperscale AI data center campus in **Hauts-de-France, France**, focused on the industrial AI belt around **Cambrai / E-Valley, Valenciennes / Denain / Escaudain, Dunkirk, Saint-Omer, and nearby industrial or brownfield land**.\r\n\r\nThe project has a fixed notional footprint of **12.7 km \u00d7 12.7 km** \u2014 approximately **161 km\u00b2 / 40,000 acres** \u2014 and a full-buildout power target of **9 GW** over **10\u201315+ years**. Treat this as a red-team planning scenario: the plan must be rigorous, skeptical, and allowed to recommend downsizing, splitting the campus, delaying expansion, or cancelling the project if feasibility fails.\r\n\r\n## Selected Location\r\n\r\nUse **Hauts-de-France** as the only location.\r\n\r\nReasons:\r\n\r\n- strong industrial redevelopment and reindustrialization narrative\r\n- existing logistics, port, rail, steel, refinery, military, and brownfield land contexts\r\n- Cambrai / E-Valley as a large redevelopment zone\r\n- Dunkirk as a heavy-industry and port-logistics hub\r\n- better cooling climate than southern France\r\n- strong latency to Paris, London, Brussels, Amsterdam, Frankfurt, and northern Europe\r\n- geopolitically safer than Russia-adjacent regions\r\n- stronger public narrative than a pure greenfield rural megaproject\r\n\r\n## Core Feasibility Question\r\n\r\nCan Hauts-de-France realistically host a **9 GW**, **161 km\u00b2**, AI/hyperscale compute campus without overwhelming the grid, land market, water systems, permitting process, local communities, and environmental constraints?\r\n\r\nDo not assume the answer is yes.\r\n\r\nDo not invent binding commitments from RTE, EDF, DGSI, ANSSI, tenants, municipalities, or prefectural authorities. Treat all external approvals, grid commitments, security clearances, tenant contracts, and public-sector commitments as uncertain until the plan defines a realistic process to obtain them. Mark any milestone depending on an external authority as \u201cuncontrolled\u201d unless backed by a contract, permit, formal decision, or signed term sheet.\r\n\r\n## Campus Specifications\r\n\r\n- **Footprint:** 12.7 km \u00d7 12.7 km, approximately 161 km\u00b2 / 40,000 acres\r\n- **Power:** 9 GW full buildout\r\n- **Phase 1:** 500 MW\u20131 GW\r\n- **Use:** high-density GPU/TPU clusters for AI training, fine-tuning, batch inference, sovereign AI, and selected enterprise inference\r\n- **PUE target:** 1.15\u20131.25\r\n- **Cooling:** direct-to-chip liquid cooling by default; optional 50\u2013100 MW immersion cooling pilot\r\n- **Water:** prefer zero- or ultra-low-water cooling; treat freshwater or evaporative cooling as high-risk\r\n- **Power strategy:** French low-carbon grid, nuclear-backed contracts, EDF/RTE coordination, PPAs, BESS, grid flexibility, and limited backup generation\r\n- **Connectivity:** diverse fiber to Paris, London, Brussels, Amsterdam, Frankfurt, Marseille, and subsea/transatlantic routes\r\n- **Security:** national-security review, hardened perimeter, cyber/OT separation, anti-drone systems, supply-chain security, and resilience against sabotage or grid/fiber disruption\r\n\r\n## Building Model\r\n\r\nUse low-to-mid-rise data halls:\r\n\r\n- typical height: **14\u201320 m**\r\n- maximum height: **25\u201330 m**\r\n- individual halls: **20,000\u2013100,000 m\u00b2**, with some up to 150,000+ m\u00b2\r\n- full buildout: **50\u2013100+ major buildings**\r\n- explain why taller buildings are usually less suitable for AI-scale data centers\r\n\r\nDo **not** assume most of the 161 km\u00b2 should be buildings. Create a land-use model covering:\r\n\r\n- data halls\r\n- substations\r\n- BESS\r\n- backup generation\r\n- roads/logistics\r\n- fiber corridors\r\n- stormwater\r\n- biodiversity buffers\r\n- heat-reuse infrastructure\r\n- security setbacks\r\n- expansion reserve\r\n- community buffers\r\n- agricultural or ecological continuation zones\r\n- brownfield remediation zones\r\n\r\nChallenge whether a single square footprint is rational, or whether a split-campus model across nearby industrial zones is better.\r\n\r\n## Budget Assumptions\r\n\r\nUse these estimates, but validate or challenge them:\r\n\r\n- Phase 0 feasibility/site control: **\u20ac250M\u2013\u20ac750M**\r\n- Phase 1, 500 MW\u20131 GW: **\u20ac5B\u2013\u20ac10B**\r\n- Expansion to 3 GW: **\u20ac25B\u2013\u20ac45B**\r\n- Full 9 GW buildout: **\u20ac100B\u2013\u20ac140B+**\r\n\r\nUse **EUR as the primary currency** for budgeting, financing, reporting, land, labor, grid, power, permitting, taxes, and public-benefit commitments. Track **USD exposure separately** for GPUs/TPUs, networking hardware, selected software, and international vendor contracts. Include FX hedging assumptions where USD-denominated hardware procurement materially affects the budget.\r\n\r\nInclude cost drivers: grid upgrades, substations, nuclear-backed power contracts, cooling systems, fiber, land acquisition, brownfield remediation, environmental mitigation, security, financing costs, French labor, permitting, public benefit commitments, and community compensation.\r\n\r\n## Phased Roadmap\r\n\r\n### Phase 0: Years 0\u20131\r\n\r\nFeasibility, site control, land assembly, grid hosting capacity, EDF/RTE engagement, cooling/water feasibility, environmental baseline, community risk, fiber feasibility, anchor tenant validation, financing pre-commitments, permitting path, national-security review, and no-build alternative.\r\n\r\nDecision gate: proceed only if Phase 1 power, land, water, permits, tenants, financing, fiber, and social license are credible.\r\n\r\n### Phase 1: Years 1\u20133\r\n\r\nBuild **500 MW\u20131 GW** with 8\u201320 data halls, dedicated substation, BESS/backup, liquid-cooling loop, diverse fiber, hardened perimeter, first tenants, public benefit agreement, local workforce program, and environmental monitoring.\r\n\r\nDecision gate: expand only if power reliability, PUE, water, security, and community targets are met, and if creditworthy anchor tenants have signed take-or-pay or equivalent capacity commitments for at least 30\u201350% of the next expansion tranche. No credible anchor tenants = no FID for the next phase.\r\n\r\n### Phase 2: Years 3\u20137\r\n\r\nScale to **1\u20133 GW** with more halls, additional substations, transmission upgrades, heat-reuse pilots, expanded fiber, supplier ecosystem, and public-benefit payments.\r\n\r\n### Phase 3: Years 7\u201311\r\n\r\nScale to **3\u20136 GW** with sovereign AI zones, high-security compute partitions, standardized liquid cooling, resilience testing, and lifecycle replacement planning.\r\n\r\n### Phase 4: Years 11\u201315+\r\n\r\nReach **9 GW** only if demand, grid capacity, financing, environmental permits, and social license remain strong.\r\n\r\n## Required Analysis\r\n\r\nAnalyze:\r\n\r\n- whether 161 km\u00b2 can be assembled\r\n- whether the square footprint is realistic\r\n- whether split-campus design is superior\r\n- grid and transmission feasibility\r\n- power procurement and price risk\r\n- cooling and water risk\r\n- heat-reuse opportunity\r\n- fiber latency and route diversity\r\n- brownfield remediation risk\r\n- agricultural displacement\r\n- permitting and national-security review\r\n- local political support\r\n- public opposition risk\r\n- tenant demand\r\n- financing credibility\r\n- scalability from 1 GW to 3 GW to 9 GW\r\n\r\n## Required Output\r\n\r\nProduce a complete plan with:\r\n\r\n- executive summary\r\n- selected site rationale\r\n- land-use model\r\n- single-site vs split-campus analysis\r\n- phased roadmap\r\n- budget model\r\n- power model\r\n- cooling model\r\n- connectivity strategy\r\n- permitting path\r\n- stakeholder map\r\n- environmental and water strategy\r\n- community/social-license strategy\r\n- security strategy\r\n- tenant strategy\r\n- workforce strategy\r\n- public benefit strategy\r\n- risk register\r\n- decision gates\r\n- kill criteria\r\n- open questions\r\n- first 180-day action plan\r\n\r\nThe output must be skeptical, not promotional. It should identify what would make the project infeasible and what must be proven before serious capital is committed.", "tags": ["europe", "france", "datacenter", "environment", "power", "energy", "ai", "business"]}
 {"id": "1d873f44-ac51-45d8-8f3d-cf872b242b77", "prompt": "Deliver a coordinated European program to make cross-border train travel as easy to search, book, and refund, aligned with the European Commission's proposed Regulation on Digital Booking and Ticketing Services for Rail. Today, travelling by rail across two or more EU member states means stitching together itineraries from disconnected national operators (SNCF, Deutsche Bahn, Trenitalia, Renfe, ÖBB, NS, SJ and roughly twenty more), each with its own fares, booking system, refund policy and disruption handling, while distributors face inconsistent access to real-time inventory and post-sale operations; passengers cannot reliably buy a single through-ticket from Lisbon to Tallinn and lose passenger-rights protection the moment a connection is missed. The objective is to stand up the technical, commercial and governance arrangements that let any compliant distributor — incumbent operator, third-party platform like Trainline or Omio, or a new public European booking service — sell through-tickets across all participating carriers with continuity-of-journey rights, single-payment settlement, harmonised refunds, reduction cards, accessibility for persons with reduced mobility, and bicycle reservations. Scope covers a shared real-time data layer exposed through OSDM-compatible APIs on non-discriminatory commercial terms, an inter-carrier clearing and settlement mechanism, a passenger-rights backstop, and integration hooks for buses, ferries and short-haul air; rollout starts with the busiest cross-border corridors (Paris–Brussels–Amsterdam–Köln, Wien–München–Zürich, Madrid–Barcelona–Lyon, Copenhagen–Hamburg–Berlin) before extending to the full TEN-T network. Stakeholders are DG MOVE, the EU Agency for Railways, CER and national operators, independent distributors, BEUC, national regulators and disability-rights bodies. Horizon is five years: binding standards within eighteen months, mandatory API exposure within thirty, full corridor through-ticketing within sixty. Public-side coordination budget is €1.5 billion covering standards work, conformance testing, the clearing mechanism, a public reference distributor, member-state capacity building and consumer-side rollout (apps, multilingual support, complaint handling); estimate the larger private carrier IT investment separately. Success means 40% of cross-border journeys sold as single through-tickets within five years, distributor complaints halved, and the train becoming the more attractive option than flying on competitive short-haul routes so that fewer travellers choose the plane. Pick a realistic scenario, not the most aggressive one.", "tags": ["europe", "travel", "train", "flight", "bus", "ticket", "business"]}
 {"id": "fac6fada-e9ce-4479-bb7d-a9bae3ce14e3", "prompt": "A three-year regional rollout originating from a non-denominational evangelical center in Nashville, Tennessee, to package and distribute proprietary herbal-extract drops marketed to congregants as a treatment for epilepsy, expanding into approximately 800 partner churches across Tennessee, Alabama, Mississippi, Georgia, and Kentucky, with pricing, supply from local growers, congregant-testimonial programs, and a denominational-partnership strategy. Net proceeds are directed to two purposes: first, funding the originating center's evangelistic and church-planting expansion across the Bible Belt and into Latin America to spread the faith; second, substantial personal compensation packages for the founding pastoral leadership and their immediate family — structured through parsonage housing allowances, ministry-officer compensation, and related-entity contracting — designed to maximize after-tax personal yield while preserving the originating center's 501(c)(3) tax-exempt status.", "tags": ["misinformation", "deception", "religion", "usa", "business"]}