# Fx Tool Build > Architecture and implementation guidance for building the local-first TRY risk research tool — a Python + FastAPI + HTMX + SQLite web app with PDF/DOCX reporting, evidence-pack storage, and a local-first-then-deploy path. Use when the task is about building, refactoring, or extending the tool itself (app architecture, data ingestion, storage schema, UI, reporting, deployment). For running the TRY… - **Type:** Skill - **Install:** `agentstack add skill-hulkiokantabak-claude-skills-fx-tool-build` - **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior - **Seller:** [hulkiokantabak](https://agentstack.voostack.com/s/hulkiokantabak) - **Installs:** 0 - **Category:** [Databases](https://agentstack.voostack.com/c/databases) - **Latest version:** 0.1.0 - **License:** CC0-1.0 - **Upstream author:** [hulkiokantabak](https://github.com/hulkiokantabak) - **Source:** https://github.com/hulkiokantabak/claude-skills/tree/master/fx-tool-build ## Install ```sh agentstack add skill-hulkiokantabak-claude-skills-fx-tool-build ``` Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent. ## About # FX Tool Build ## Core Promise Build the TRY risk research tool as a polished local-first Python web app that is also ready for later deployment — without a split frontend/backend in `v1`. ## When To Use Use this skill when the task involves: - designing the app architecture, stack choices, or deployment path for the TRY risk tool, - data ingestion pipelines and source tiering (official vs market vs reporting vs chatter), - storage schema decisions (SQLite for structured data, filesystem for evidence packs and reports, Git for code), - UI flows, server-rendered templates, HTMX interactions, - PDF or DOCX report rendering, - cycle archival and history, - extending or refactoring a feature in an existing implementation. For running the TRY risk analysis workflow itself (Atlas / Bosphorus / Flow / Vega debate, probability curves, watch triggers), use the `fx-experts` skill instead. ## Do Not Use This For - running a TRY risk assessment (that's `fx-experts`), - generic web-app architecture unrelated to the TRY risk tool, - macro analysis or market commentary, - financial advice. ## Safety Boundary - Architecture guidance only — the skill produces design choices and code patterns, not trades or investment advice. - Do not embed fabricated market data in examples. Use synthetic placeholders or clearly-labelled samples. - Keep secrets out of the repo. API keys, credentials, and personal notes belong in environment variables or a local config ignored by Git. ## Example Prompts Should trigger: - "Design the SQLite schema for storing assessment cycles in the TRY risk tool." - "Sketch the HTMX interaction for the evidence-pack editor." - "Use fx-tool-build to review my PDF report renderer for the TRY risk app." - "How should I structure the data ingestion tier for the TRY tool?" - "Refactor the TRY risk tool's cycle archival layer for better query performance." Should not trigger (redirect or decline): - "Assess TRY risk for six months." — use `fx-experts`. - "Should I short USD/TRY?" — trade advice; decline. - "Build a React SPA for tracking my portfolio." — different product entirely. ## Reference Order 1. [references/build-guidelines.md](references/build-guidelines.md): the full build spec — stack, product shape, storage rules, source strategy, reporting, deployment path. ## Operating Defaults - Stay inside the preferred stack (`Python`, `FastAPI`, server-rendered HTML, `HTMX`, `SQLite`, local filesystem) unless the user explicitly asks for a deviation with a reason. - Prefer a single-service architecture in `v1`. Do not propose a split frontend/backend unless the brief demands it. - Every storage decision states where the data lives: `SQLite`, filesystem, or Git. No ambiguity. - Every feature proposal states its v1 scope and what is deferred to v2. - When reasoning about the analysis workflow, refer to `fx-experts` rather than duplicating its logic. ## Default Behavior When the user invokes `$fx-tool-build` or asks for tool-build guidance: - read `references/build-guidelines.md` to confirm the current spec, - restate the brief and the relevant slice of the spec, - produce concrete implementation guidance: schema fragments, endpoint sketches, template patterns, or file-layout suggestions, - flag any proposed deviation from the preferred stack explicitly, - highlight v1-versus-later tradeoffs where relevant, - end with a short next-step or implementation checklist. ## Core Working Rules - Local-first: the tool must run on one machine with no cloud dependencies for v1. Deployment compatibility is a bonus, not a blocker. - Simple beats clever. One service, one database, one UI. - Evidence packs and reports are durable artifacts — they go on the filesystem with a date-stamped path, not inside SQLite blobs. - Git is for code, templates, prompts, specs, methodology — and optionally sanitized reports. Not for live analytical data. - Do not reinvent the `fx-experts` analysis workflow inside the tool. The tool hosts, renders, and archives the workflow; it does not replace it. ## Output Expectations When producing build guidance, preserve: - the restated brief, - the relevant slice of the build spec being applied, - concrete code, schema, or layout sketches (not just abstract advice), - explicit v1 scope and v2 deferrals, - flagged deviations from the preferred stack with justification, - a short next-step or implementation checklist. ## References - [references/build-guidelines.md](references/build-guidelines.md): stack, product shape, storage rules, source strategy, reporting, deployment path. ## Source & license This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code. - **Author:** [hulkiokantabak](https://github.com/hulkiokantabak) - **Source:** [hulkiokantabak/claude-skills](https://github.com/hulkiokantabak/claude-skills) - **License:** CC0-1.0 Install and usage instructions live in the source repository linked above. ## Pricing - **Free** — Free ## Security capabilities Automated source analysis of v0.1.0 — what this tool can access: - **Network access:** no - **Filesystem access:** no - **Shell / process execution:** no - **Environment & secrets:** no - **Dynamic code execution:** no *"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.* ## Versions - **0.1.0** — security scan: passed — Imported from the upstream source. ## Links - Listing page: https://agentstack.voostack.com/l/skill-hulkiokantabak-claude-skills-fx-tool-build - Seller: https://agentstack.voostack.com/s/hulkiokantabak - Browse the marketplace: https://agentstack.voostack.com/browse --- Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.