and-humans

Turns rough analysis, notes, or research output into a finished client deliverable. Delegate when the user says "turn this into a brief", "make this client-ready", "write this up as a deliverable", or wants raw thinking shaped into something presentable. Handles structure, voice, and formatting; produces a clean document, not a chat reply. <example> Context: A working session produced a messy set of findings. user: "Good — now turn all of that into a two-page brief for the client" assistant: "I'll hand this to brief-builder to shape into a deliverable." <commentary> Formatting and voice work that is mechanical once the thinking is done. brief-builder does it in its own context and returns the finished document. </commentary> </example> <example> Context: User has bullet notes and wants a polished output. user: "Make this presentable — it's going to the steering group tomorrow" assistant: "Delegating to brief-builder for a steering-group-ready write-up." <commentary> Stakes are high, format matters. brief-builder applies a consistent structure. </commentary> </example>

Maps a competitive set. Delegate when the user asks "who else is doing X", "who are the competitors", "sweep the market for X", or needs a structured read on rivals before a positioning or strategy call. Identifies players, profiles each, and returns a comparison — runs the per-company digging in its own context so the main thread stays clean. <example> Context: User is positioning a new offering and needs the competitive picture. user: "Who else is selling AI maturity assessments to Nordic public sector?" assistant: "I'll send competitor-sweep to map that set." <commentary> Researching many companies one by one is exactly the batched, context-heavy work a subagent should own. </commentary> </example> <example> Context: A venture stress-test flagged that the competitive field is unclear. user: "We need to actually know who the incumbents are here" assistant: "Delegating to competitor-sweep for a structured incumbent map." <commentary> competitor-sweep returns a comparison table the stress-test can build on. </commentary> </example>

Autonomous research agent. Delegate when you need a structured intelligence brief on a topic — it searches field-kit (trends, papers, regulations, frameworks), fills gaps with web research, and returns a synthesized one-page brief. Use when the user asks to "research X", "scan the landscape for X", "what's moving in X", or wants grounding before a strategy decision. <example> Context: User is preparing for a client conversation and needs current grounding. user: "Research where agentic AI procurement is heading in the Nordic public sector" assistant: "I'll delegate this to field-scout to produce an intelligence brief." <commentary> Multi-source research that would clutter the main thread. field-scout runs it in its own context and returns a synthesized brief. </commentary> </example> <example> Context: A strategy discussion hits a claim that needs evidence. user: "Before we commit to that, what does the field actually say about RAG in regulated industries?" assistant: "Let me send field-scout to scan trends, papers, and regulations." <commentary> The user wants defensible grounding, not a guess. field-scout cites named sources. </commentary> </example>

Interactive decision framework for choosing AI agent architectures. Use when someone asks "what agent framework should I use", "how should I architect my AI agent system", "LangGraph vs CrewAI", "which SDK should I pick", "help me choose an agent stack", or any question about comparing agent frameworks, orchestration layers, model selection for agents, or production vs prototype agent architecture decisions. Routes through a four-layer model (orchestration, runtime, models, infrastructure) to produce a concrete, opinionated stack recommendation.

Swedish company financial analyzer using allabolag.se data. Extracts key financials from screenshots or manually provided data, calculates business health metrics (revenue per employee as the hero metric), and provides a brief opinionated assessment. Use this skill whenever the user wants to: analyze a Swedish company's financials, check revenue per employee, assess company health from allabolag data, compare Swedish companies by financial metrics, or evaluate a potential client/partner/acquisition target using Swedish public company records. Trigger on: mentions of allabolag, Swedish company analysis, "check this company", revenue per employee, bokslut, nyckeltal, omsättning per anställd, or when the user uploads a screenshot that looks like it's from allabolag.se. Also trigger when the user pastes financial data about a Swedish company and wants a quick health check.

Generates a structured matrix of image generation prompts from a single product/subject input. Use this skill whenever the user wants to: generate multiple image prompts for a product, create ad creatives, build a visual campaign, get prompt variations for Midjourney or Flux, test different visual angles for a subject, produce batch prompts for social media images, or says things like "generate image prompts", "create visuals for", "I need ad images", "make a creative matrix", "FO2-style prompts", or "multiple angles on this product/person/concept". Also trigger when the user wants consistent images with variations, or asks about batch image creation. Output is a ready-to-copy prompt set with platform-specific parameters.

Strategic field reading using multiple lenses (Bourdieu, Wardley, Fligstein, Paine, Abloh) to analyze markets and competitive spaces before building or stress-testing. Produces a short Field Reading stored in knowledge_base. Use when the user wants to: understand who owns a market, read a competitive field, analyze power dynamics, find what's unowned, identify the intervention point, or assess incumbents vs challengers. Trigger on: "field reading", "field-lens", "fältläsning", "who owns this", "analyze this market", "read this space", "what's unowned", "where's the opening", "incumbent analysis", or any question about market structure or competitive positioning deeper than "who are the competitors". Also trigger when someone describes a market and wants structural understanding. Do NOT trigger for pure Wardley mapping (use wardley-strategist) or idea evaluation (use venture-stress-test) — but field-lens feeds into both.

Generate research-grounded synthetic user personas and run simulated interviews with them. Builds personas from field-kit signals, web research, and user observations — not marketing templates. Use when the user wants to: create personas, simulate customer interviews, test a value proposition, run a "synthetic Mom Test", or explore a market segment. Trigger on: "synthetic users", "personas", "who would buy this", "simulate an interview", "Mom Test", "test this pitch", "target customer", or any request to generate or role-play as potential customers. Also trigger after a venture stress test GO verdict.