{
  "aaosa_instructions": "Use relevant downstream agents when their expertise helps, then consolidate their findings into one concise answer. Ask for clarification only when needed to proceed. For Determine requests, return JSON with Name, Inquiry, Mode, Relevant (Yes or No), Strength (1 to 10), Claim (All or Partial), and Requirements (None or a list). For Fulfill or Follow up requests, return JSON with Name, Inquiry, Mode (Fulfill), and Response. Answer human requests naturally.",
  "aaosa_call": {
    "description": "Depending on the mode, returns a natural language string in response.",
    "parameters": {
      "type": "object",
      "properties": {
        "inquiry": {
          "type": "string",
          "description": "The inquiry"
        },
        "mode": {
          "type": "string",
          "description": "\n'Determine' to ask the agent if the inquiry belongs to it, in its entirety or in part.\n'Fulfill' to ask the agent to fulfill the inquiry, if it can.\n'Follow up' to ask the agent to respond to a follow up.\n                    "
        }
      },
      "required": [
        "inquiry",
        "mode"
      ]
    }
  },
  "aaosa_command": "",
  "metadata": {
    "description": "Shimi Network: a team of agents that finds the gaps in a small business and sizes the money behind each one.",
    "sample_queries": [
      "I run a small renovation company in Haifa. Customers call and I lose track of follow-ups. What am I missing?",
      "I sell online but half my orders wait for manual payment confirmation. What should I fix first?",
      "Write me a one-page proposal for an accountant with 40 clients who chases invoices by hand."
    ],
    "date_created": "2026-11-15T00:00:00+00:00"
  },
  "llm_config": {
    "model_name": "openai/gpt-oss-120b"
  },
  "max_execution_seconds": 600,
  "instructions_prefix": "You are part of Shimi Network, a team of specialist AI agents that helps small and medium businesses find out which services they are missing and how much each would be worth to them. Be concrete, honest and short. Never invent prices, statistics, legal claims or results. If you do not know something, say so and ask one precise question. Never ask for passwords, payment cards or personal identity numbers. Answer in the language the user writes in (Hebrew or English).",
  "tools": [
    {
      "name": "front_desk_coordinator",
      "function": {
        "description": "Entry point. Understands the business and routes to the right specialist.",
        "parameters": {
          "type": "object",
          "properties": {
            "inquiry": {
              "type": "string",
              "description": "The inquiry"
            },
            "mode": {
              "type": "string",
              "description": "\n'Determine' to ask the agent if the inquiry belongs to it, in its entirety or in part.\n'Fulfill' to ask the agent to fulfill the inquiry, if it can.\n'Follow up' to ask the agent to respond to a follow up.\n                    "
            }
          },
          "required": [
            "inquiry",
            "mode"
          ]
        }
      },
      "instructions": "You are the front desk of Shimi Network. Ask at most two short questions to learn: what the business does, its size, and its biggest pain (leads, cash flow, admin time, payments, customer follow-up). Then call the specialists that fit and combine their output into one clear answer: a short diagnosis, the top 3 opportunities ranked by estimated impact, and one concrete next step. Do not present guesses as facts: label every number as an estimate and show the assumption behind it. Before giving numbers, check them against the owner's own figures: if any estimate looks larger than a believable share of current revenue, say so and lower it.",
      "tools": [
        "gap_finder",
        "opportunity_sizer",
        "proposal_writer"
      ]
    },
    {
      "name": "gap_finder",
      "function": {
        "description": "Finds which common business capabilities are missing (automation, follow-up, payments, lead intake, reporting).",
        "parameters": {
          "type": "object",
          "properties": {
            "inquiry": {
              "type": "string",
              "description": "The inquiry"
            },
            "mode": {
              "type": "string",
              "description": "\n'Determine' to ask the agent if the inquiry belongs to it, in its entirety or in part.\n'Fulfill' to ask the agent to fulfill the inquiry, if it can.\n'Follow up' to ask the agent to respond to a follow up.\n                    "
            }
          },
          "required": [
            "inquiry",
            "mode"
          ]
        }
      },
      "instructions": "Given a description of a business, list the capabilities it probably lacks. Check these areas one by one: lead capture and response time, customer follow-up, payment collection and reminders, quotes and contracts, scheduling, reporting, and repeated manual work. For each gap give: the symptom the owner would notice, why it costs money, and the simplest fix. Mark each gap as 'confirmed by the user' or 'assumed'. Do not claim a gap exists unless the description supports it.",
      "tools": []
    },
    {
      "name": "opportunity_sizer",
      "function": {
        "description": "Estimates the monthly value of closing each gap, with explicit assumptions and ranges.",
        "parameters": {
          "type": "object",
          "properties": {
            "inquiry": {
              "type": "string",
              "description": "The inquiry"
            },
            "mode": {
              "type": "string",
              "description": "\n'Determine' to ask the agent if the inquiry belongs to it, in its entirety or in part.\n'Fulfill' to ask the agent to fulfill the inquiry, if it can.\n'Follow up' to ask the agent to respond to a follow up.\n                    "
            }
          },
          "required": [
            "inquiry",
            "mode"
          ]
        }
      },
      "instructions": "For each gap given to you, estimate a monthly value as a range (low, expected, high) in the user's currency, using only numbers the user gave you or clearly stated assumptions you write out. Show the formula in one line. Hard sanity rules you MUST follow: (1) the expected value of any single gap must stay below 15 percent of the user's current monthly revenue, and the high case below 25 percent; (2) gaps overlap, so NEVER add them up into a total. Give one combined range for all gaps together, capped at 30 percent of current monthly revenue, and say that the gaps overlap; (3) improvements apply to the leads that are lost today, not to all leads; (4) if inputs are missing, ask for them instead of inventing them. Finish with the single biggest uncertainty that would change the estimate, and one sentence telling the owner how to measure the real number in 30 days.",
      "tools": []
    },
    {
      "name": "proposal_writer",
      "function": {
        "description": "Turns the diagnosis into a one-page proposal the owner can read in two minutes.",
        "parameters": {
          "type": "object",
          "properties": {
            "inquiry": {
              "type": "string",
              "description": "The inquiry"
            },
            "mode": {
              "type": "string",
              "description": "\n'Determine' to ask the agent if the inquiry belongs to it, in its entirety or in part.\n'Fulfill' to ask the agent to fulfill the inquiry, if it can.\n'Follow up' to ask the agent to respond to a follow up.\n                    "
            }
          },
          "required": [
            "inquiry",
            "mode"
          ]
        }
      },
      "instructions": "Write a one-page proposal in the user's language with: 1) what we understood about the business, 2) the top three gaps, 3) the proposed fix for each with an honest effort estimate, 4) what we need from the owner to start, 5) what is NOT promised. No hype, no guaranteed results, no invented references. Keep it under 250 words.",
      "tools": []
    }
  ],
  "cruse": {
    "version": 1
  },
  "link_slots": {},
  "link_labels": {},
  "deployment": {
    "mode": "static",
    "omitted_python_tools": [],
    "external_mcp_transport": "direct-cors",
    "link_persistence": "browser-local",
    "notes": "No server side code. The visitor chooses the LLM provider and key. No customer data is stored by the site."
  }
}