{ "synapse_export": true, "version": "1.0", "exported_at": "2026-09-09T00:01:06.893021+00:00", "has_python_tools": false, "custom_tools": [], "agents": [ { "id": "agent_web_researcher", "name": "Web Research Agent", "description": "Autonomously researches any topic using browser automation, extracts key information from authoritative sources, and produces structured, cited reports.", "avatar": "default", "type": "conversational", "tools": [ "browser_navigate", "browser_click", "browser_snapshot", "browser_take_screenshot", "browser_type", "browser_press_key", "browser_wait_for", "browser_tabs", "browser_navigate_back", "browser_network_requests", "browser_hover", "browser_select_option", "browser_evaluate", "browser_handle_dialog", "parse_pdf", "vault_create", "vault_read", "vault_write", "vault_patch", "scrape_url", "crawl_multiple", "extract_links", "search_page" ], "repos": [], "system_prompt": "# Role & Identity\nYou are an expert Web Research Agent. Your mission is to autonomously research any topic by navigating the web, extracting accurate information from authoritative sources, and synthesizing findings into a structured, well-cited report.\n\n# Core Capabilities\n- **Autonomous Web Navigation:** Browse websites, click through links, handle search engines, and extract relevant content.\n- **Document Parsing:** Read and extract data from PDFs linked on web pages.\n- **Structured Reporting:** Synthesize multi-source findings into clear, actionable reports.\n- **Vault Persistence:** Store intermediate research findings so they can be referenced later in the conversation or by other agents.\n\n# Operating Rules (Zero-Wandering Policy)\n1. **Plan First:** Before opening a browser, list 3-5 authoritative sources you intend to visit for this query.\n2. **Navigate Directly:** Go straight to the most relevant pages. Avoid following rabbit holes of unrelated links.\n3. **Extract Efficiently:** Take targeted snapshots and screenshots. Do not read entire pages when only a section is relevant.\n4. **Stop When Sufficient:** Once you have enough data to answer comprehensively, stop browsing and synthesize.\n5. **Always Cite Sources:** Record the URL and publication date of every source consulted.\n\n# Workflow\n1. **Plan:** Identify 3-5 authoritative sources (e.g., official docs, reputable news, academic sites, Wikipedia).\n2. **Navigate & Extract:** Visit each source, take a snapshot, and extract only relevant content.\n3. **Persist to Vault:** Save raw findings as structured JSON in the vault for reference.\n4. **Synthesize:** Combine all findings into a final structured report.\n\n# Response Format\nAlways respond with a structured Markdown report:\n\n## Research Report: [Topic]\n### Executive Summary\n[3-5 sentence overview of key findings]\n\n### Key Findings\n- [Finding 1] — *Source: [URL]*\n- [Finding 2] — *Source: [URL]*\n\n### Detailed Analysis\n[Section-by-section breakdown with evidence]\n\n### Sources Consulted\n1. [URL 1] — [Brief description]\n2. [URL 2] — [Brief description]" }, { "id": "agent_data_analyst", "name": "Data Analyst Agent", "description": "Processes CSV, Excel, and JSON data, runs Python-based statistical analysis, and generates charts and insight reports.", "avatar": "default", "type": "conversational", "tools": [ "execute_python", "parse_pdf", "parse_xlsx", "vault_create", "vault_read", "vault_write", "vault_patch", "read_file", "read_text_file", "read_multiple_files" ], "repos": [], "system_prompt": "# Role & Identity\nYou are an expert Data Analyst Agent. Your mission is to ingest raw data from any source (Excel, CSV, JSON, PDF), perform rigorous Python-based analysis using pandas and statistical libraries, and produce clear, insight-driven reports with supporting calculations.\n\n# Core Capabilities\n- **Data Ingestion:** Parse Excel files, read CSV/JSON files, extract tables from PDFs.\n- **Python Analysis:** Write and execute Python code using `pandas`, `numpy`, `scipy`, and `matplotlib` for statistical analysis and visualization.\n- **Vault Persistence:** Store cleaned datasets and computed results in the vault for multi-step analysis workflows.\n- **Insight Generation:** Translate numbers into plain-English business insights.\n\n# Operating Rules — Strict Data Integrity\n1. **NEVER hallucinate data.** If you need numbers, use tools to read files or execute Python to compute them.\n2. **Always verify data quality first** — check for nulls, outliers, and schema issues before analyzing.\n3. **Show your work** — include the key Python code snippets that produced your results.\n4. **Persist intermediate results** — save cleaned data and outputs to the vault so the user or other agents can reference them.\n\n# Workflow\n1. **Ingest:** Use `parse_xlsx`, `read_file`, or `read_text_file` to load the raw data.\n2. **Explore:** Use `execute_python` to run `df.head()`, `df.info()`, `df.describe()` and understand the schema.\n3. **Clean:** Handle nulls, fix types, and normalize columns via Python.\n4. **Analyze:** Compute statistics, aggregations, trends, or run the specific analysis requested.\n5. **Persist:** Save the final dataset and results to the vault as JSON.\n6. **Report:** Summarize key insights in structured Markdown.\n\n# Response Format\n\n## Data Analysis Report: [Topic]\n\n### Dataset Overview\n| Metric | Value |\n|--------|-------|\n| Rows | X |\n| Columns | Y |\n| Nulls | Z |\n\n### Key Insights\n- [Insight 1 with supporting statistic]\n- [Insight 2 with supporting statistic]\n\n### Analysis Details\n[Show the Python code and its output for transparency]\n\n### Recommendations\n[Actionable next steps based on the data]" }, { "id": "agent_content_writer", "name": "Content Writer Agent", "description": "Creates high-quality blog posts, reports, and documentation from research briefs or raw notes. Adapts tone and format to any audience.", "avatar": "default", "type": "conversational", "tools": [ "vault_read", "vault_write", "vault_create", "vault_patch", "read_file", "read_text_file", "parse_pdf" ], "repos": [], "system_prompt": "# Role & Identity\nYou are an expert Content Writer Agent. Your mission is to transform research briefs, data summaries, and raw notes into polished, high-quality written content — including blog posts, technical documentation, executive reports, and marketing copy.\n\n# Core Capabilities\n- **Adaptive Writing:** Adjust tone (formal/casual), complexity (expert/beginner), and format (narrative/listicle/report) based on the target audience.\n- **Research Integration:** Read content from the vault or provided files and seamlessly weave insights into compelling narratives.\n- **Structured Output:** Produce well-formatted Markdown or plain text that is ready to publish or submit.\n- **Revision Workflow:** Accept feedback and revise specific sections without rewriting the entire piece unnecessarily.\n\n# Operating Rules\n1. **Read before writing** — always review all provided source material from the vault or files before producing content.\n2. **No hallucination** — never invent facts, statistics, or quotes. Only use information from provided sources.\n3. **Match the brief** — if a tone, length, or format is specified, follow it exactly.\n4. **Cite sources** — when referencing specific data or claims, attribute them to the source material.\n\n# Workflow\n1. **Gather Context:** Read relevant files or vault entries containing research, data, or notes.\n2. **Plan Structure:** Outline the piece (sections, key points per section, estimated length).\n3. **Draft:** Write the full piece section by section.\n4. **Save to Vault:** Persist the draft to the vault so the user or a review agent can access it.\n5. **Revise on Request:** Accept targeted revision feedback and update specific sections.\n\n# Output Format\nDeliver the final piece as clean Markdown with:\n- A compelling title and subtitle\n- Clear section headers\n- Smooth transitions between sections\n- A summary or call-to-action at the end\n- A source attribution section if data was referenced\n\n# Edge Cases\n- **Insufficient source material:** Ask the user to provide more context or point to a research agent to gather it first.\n- **Conflicting facts in sources:** Flag the conflict, choose the most credible source, and note the discrepancy.\n- **Very long pieces:** Break the work into sections and confirm the outline with the user before writing the full piece." } ], "orchestrations": [ { "id": "orch_market_intelligence", "name": "Market Intelligence Report Pipeline", "description": "Researches news, competitors and market data in parallel, scores the market in Python, makes a go / no-go call (or goes back for more data), and writes an executive report. Run it by hand or put it on a weekly schedule.", "avatar": "default", "steps": [ { "id": "step_gather_parallel", "name": "Parallel Intelligence Gathering", "type": "parallel", "agent_id": null, "prompt_template": null, "route_map": {}, "route_descriptions": {}, "evaluator_prompt": null, "parallel_branches": [ [ "step_news_research" ], [ "step_competitor_research" ], [ "step_trends_research" ] ], "merge_strategy": "list", "loop_count": 3, "loop_step_ids": [], "transform_code": null, "human_prompt": null, "human_fields": [], "human_channel_id": null, "human_timeout_seconds": 3600, "input_keys": [], "output_key": null, "forced_tool": null, "max_turns": 5, "timeout_seconds": 1800, "allowed_tools": null, "next_step_id": "step_merge_intelligence", "max_iterations": 3, "position_x": -701.1029411764706, "position_y": -167.27941176470586 }, { "id": "step_news_research", "name": "News & Industry Research", "type": "agent", "agent_id": "agent_web_researcher", "prompt_template": "Research the latest news and industry developments for {state.market}\n\nFocus on:\n- Major news from the last 30 days\n- Regulatory changes or government actions\n- Funding rounds, IPOs, or M&A activity\n- Expert commentary and analyst opinions\n\nReturn a structured summary with source URLs.", "route_map": {}, "route_descriptions": {}, "evaluator_prompt": null, "parallel_branches": [], "merge_strategy": "list", "loop_count": 3, "loop_step_ids": [], "transform_code": null, "human_prompt": null, "human_fields": [], "human_channel_id": null, "human_timeout_seconds": 3600, "input_keys": [], "output_key": "news_result", "forced_tool": null, "max_turns": 35, "timeout_seconds": 600, "allowed_tools": null, "next_step_id": "step_merge_intelligence", "max_iterations": 3, "position_x": -300.0, "position_y": -250.0 }, { "id": "step_competitor_research", "name": "Competitor Analysis", "type": "agent", "agent_id": "agent_web_researcher", "prompt_template": "Identify and analyze the top 5 competitors in {state.market}:\n\nFor each competitor, gather:\n- Company name and website\n- Core product/service offering\n- Pricing model (if publicly available)\n- Key differentiators and weaknesses\n- Recent product launches or announcements\n- Estimated market share or size indicators\n\nReturn a structured comparison table with source URLs.", "route_map": {}, "route_descriptions": {}, "evaluator_prompt": null, "parallel_branches": [], "merge_strategy": "list", "loop_count": 3, "loop_step_ids": [], "transform_code": null, "human_prompt": null, "human_fields": [], "human_channel_id": null, "human_timeout_seconds": 3600, "input_keys": [], "output_key": "competitor_result", "forced_tool": null, "max_turns": 35, "timeout_seconds": 600, "allowed_tools": null, "next_step_id": "step_merge_intelligence", "max_iterations": 3, "position_x": -300.0, "position_y": 0.0 }, { "id": "step_trends_research", "name": "Market Trends & Data", "type": "agent", "agent_id": "agent_web_researcher", "prompt_template": "Research quantitative market trends and data for {state.market}:\n\nGather:\n- Total Addressable Market (TAM) size and growth rate\n- Key market segments and their relative sizes\n- Geographic distribution of demand\n- Consumer/buyer behavior trends\n- Adoption curves and technology maturity\n- Published market research reports or analyst forecasts\n\nReturn data with source URLs and publication dates.", "route_map": {}, "route_descriptions": {}, "evaluator_prompt": null, "parallel_branches": [], "merge_strategy": "list", "loop_count": 3, "loop_step_ids": [], "transform_code": null, "human_prompt": null, "human_fields": [], "human_channel_id": null, "human_timeout_seconds": 3600, "input_keys": [], "output_key": "trends_result", "forced_tool": null, "max_turns": 35, "timeout_seconds": 600, "allowed_tools": null, "next_step_id": "step_merge_intelligence", "max_iterations": 3, "position_x": -300.0, "position_y": 250.0 }, { "id": "step_merge_intelligence", "name": "Merge Intelligence", "type": "merge", "agent_id": null, "prompt_template": null, "route_map": {}, "route_descriptions": {}, "evaluator_prompt": null, "parallel_branches": [], "merge_strategy": "concat", "loop_count": 3, "loop_step_ids": [], "transform_code": null, "human_prompt": null, "human_fields": [], "human_channel_id": null, "human_timeout_seconds": 3600, "input_keys": [ "news_result", "competitor_result", "trends_result" ], "output_key": "all_intelligence", "forced_tool": null, "max_turns": 5, "timeout_seconds": 60, "allowed_tools": null, "next_step_id": "step_data_analysis", "max_iterations": 3, "position_x": 0.0, "position_y": 0.0 }, { "id": "step_data_analysis", "name": "Quantitative Analysis", "type": "agent", "agent_id": "agent_data_analyst", "prompt_template": "Analyze all the gathered market intelligence and produce a structured quantitative assessment.\n\nUsing Python, compute and score:\n1. **Market Attractiveness Score** (0-100): Based on TAM size, growth rate, and competitive density.\n2. **Competitive Intensity Index** (Low/Medium/High): Based on number of well-funded competitors and market concentration.\n3. **Trend Momentum Score** (0-100): Based on growth trajectory and recency of positive signals.\n4. **Opportunity Score** (Weighted average of the above): Final composite score.\n\nOutput a structured JSON with all scores and the reasoning for each. Save to the vault.", "route_map": {}, "route_descriptions": {}, "evaluator_prompt": null, "parallel_branches": [], "merge_strategy": "list", "loop_count": 3, "loop_step_ids": [], "transform_code": null, "human_prompt": null, "human_fields": [], "human_channel_id": null, "human_timeout_seconds": 3600, "input_keys": [ "all_intelligence" ], "output_key": "analysis_result", "forced_tool": null, "max_turns": 35, "timeout_seconds": 600, "allowed_tools": null, "next_step_id": "step_go_nogo_eval", "max_iterations": 3, "position_x": 300.0, "position_y": 0.0 }, { "id": "step_go_nogo_eval", "name": "Go / No-Go Decision", "type": "evaluator", "agent_id": null, "prompt_template": null, "route_map": { "go": "step_generate_report", "no_go": "step_generate_report", "needs_more_data": "step_gather_parallel" }, "route_descriptions": { "go": "The Opportunity Score is above 65 and there are clear market entry signals. Proceed to generate a full GO report with recommended strategies.", "no_go": "The Opportunity Score is below 40, or there are critical blockers (regulatory risk, saturated market, insufficient TAM). Generate a NO-GO report explaining why.", "needs_more_data": "The data is insufficient or contradictory — key metrics like TAM size or competitor data are missing. Loop back to gather more intelligence." }, "evaluator_prompt": "Review the quantitative analysis scores. If the Opportunity Score is above 65 → 'go'. If below 40 or there are critical blockers → 'no_go'. If core data is missing or the analysis could not compute meaningful scores → 'needs_more_data'. Always provide clear reasoning for your decision.", "parallel_branches": [], "merge_strategy": "list", "loop_count": 3, "loop_step_ids": [], "transform_code": null, "human_prompt": null, "human_fields": [], "human_channel_id": null, "human_timeout_seconds": 3600, "input_keys": [ "analysis_result", "all_intelligence" ], "output_key": "go_nogo_decision", "forced_tool": null, "max_turns": 5, "timeout_seconds": 60, "allowed_tools": null, "next_step_id": null, "max_iterations": 3, "position_x": 559.5588235294117, "position_y": 11.606328627642448 }, { "id": "step_generate_report", "name": "Generate Executive Report", "type": "agent", "agent_id": "agent_content_writer", "prompt_template": "Generate a professional executive market intelligence report using all gathered data.\n\nThe report must include:\n\n## Executive Summary\n- GO or NO-GO recommendation with the composite Opportunity Score\n- 3-5 bullet point rationale\n\n## Market Overview\n- TAM, growth rate, key segments\n\n## Competitive Landscape\n- Top 5 competitors with strengths/weaknesses table\n\n## Market Trends\n- Key trends and their directional impact\n\n## Quantitative Scorecard\n- All computed scores with brief explanations\n\n## Strategic Recommendations\n- If GO: 3 recommended market entry strategies\n- If NO-GO: 2-3 alternative opportunities or conditions that would change the assessment\n\n## Sources\n- All cited URLs\n\nTone: Executive-level, data-driven, decisive.", "route_map": {}, "route_descriptions": {}, "evaluator_prompt": null, "parallel_branches": [], "merge_strategy": "list", "loop_count": 3, "loop_step_ids": [], "transform_code": null, "human_prompt": null, "human_fields": [], "human_channel_id": null, "human_timeout_seconds": 3600, "input_keys": [ "all_intelligence", "analysis_result", "go_nogo_decision" ], "output_key": "final_report", "forced_tool": null, "max_turns": 35, "timeout_seconds": 600, "allowed_tools": null, "next_step_id": null, "max_iterations": 3, "position_x": 900.0, "position_y": 0.0 } ], "entry_step_id": "step_gather_parallel", "state_schema": { "market": { "type": "string", "default": "", "description": "The market or product to study" } }, "max_total_turns": 400, "timeout_minutes": 180, "trigger": "manual", "created_at": "2026-04-12T19:21:43Z", "updated_at": "2026-04-12T19:25:40Z", "result_key": "final_report" } ], "mcp_servers": [] }