agent-helper-scripts

Diagnose and add "juice" — game feel, feedback, and tactile polish — to games, prototypes, and interactive software. Use this skill whenever the user is building, polishing, prototyping, or reviewing a game (especially action, arcade, platformer, shooter, or rhythm games), or when they ask about game feel, juice, screen shake, hit feedback, particles, easing, tweening, screen pause/hitstop, juicy effects, "making it feel good", or why their prototype feels limp, lifeless, dry, or cheap. Also trigger when reviewing UI prototypes that need tactile response, or when the user mentions Jonasson, Purho, Vlambeer, Nuclear Throne, screen shake, or Steve Swink's "Game Feel" book. Apply even to non-game interactive software when the user wants buttons, transitions, or interactions to feel responsive and satisfying.

Visual strategy, art direction, and design critique for websites. Use this skill whenever the user asks for visual direction, page design strategy, design critique, hierarchy planning, concept exploration, or art direction for any web page or site — including landing pages, service/transactional pages, dashboards, editorial layouts, portfolios, campaign pages, and design-system foundations. Also trigger when the user asks to evaluate an existing web design, generate concept territories, plan visual hierarchy, choose type/colour/image strategy, or review a design for accessibility and communication clarity. This skill produces design rationale, concept directions, hierarchy maps, system strategies, and review notes — not HTML/CSS code. If the user needs implementation, hand off to the frontend-design skill after the design direction is set.

Local-first Ansible testing for roles and modules within collections. Use whenever the user wants to add, run, scaffold, or debug tests for an Ansible collection, role, or module. Triggers include: "test my Ansible role", "add Molecule tests", "run ansible-test", "set up integration tests", "scaffold a test scenario", "run sanity checks", "write unit tests for a module", "test a collection locally". Always prefer this skill over ad hoc shell suggestions when Ansible testing is the subject. The skill covers the full testing stack: Molecule + Podman for roles, ansible-test for modules and collections (sanity, unit, integration). Python 3.12+, collections layout, and Podman are assumed throughout.

Configure and use Biome (biomejs) for TypeScript linting and formatting. Use when setting up Biome in a project, configuring lint rules, migrating from ESLint/Prettier, fixing lint errors, setting up CI pipelines with Biome, or configuring git hooks for code quality. Covers biome.json configuration, file inclusion/exclusion patterns, rule overrides, and integration with build tooling.

Conduct thorough, actionable code reviews that catch real problems without drowning in noise

Generate, modify, or audit `.codescene/code-health-rules.json` files that control CodeScene's code health scan behaviour. Use this skill whenever the user wants to customize CodeScene rule weights, disable specific smells, adjust metric thresholds, scope rules to test vs application code, apply language-specific overrides, or add in-source `@codescene` directives. Also trigger when the user asks why a CodeScene rule is firing, or wants to suppress a smell across a repo or folder subtree.

Write and maintain self-contained ExecPlans (execution plans) that a novice can follow end-to-end; use when planning or implementing non-trivial repo changes.

Using the Firecrawl MCP server to scrape, search, crawl, extract, and browse the web. Use this skill whenever the Firecrawl MCP tools are available and you need to retrieve web content, discover URLs on a site, search the web with full-page content retrieval, extract structured data from pages, perform autonomous multi-source web research, or interact with web pages through a remote browser sandbox. Trigger this skill for any task involving firecrawl_scrape, firecrawl_search, firecrawl_map, firecrawl_crawl, firecrawl_extract, firecrawl_agent, or firecrawl_browser_* tools. Also trigger when the user asks you to "scrape", "crawl", "map a site", "extract data from a page", "search with Firecrawl", or "use the browser sandbox", even if they don't mention Firecrawl by name — provided the MCP tools are connected.

Workspace-first GrepAI search workflow for programming agents (v0.34+): project-scoped semantic search, token-efficient output, and reliable fallbacks.

Design, implement, and maintain applications using hexagonal architecture (ports and adapters). Use when (1) designing new systems requiring clear separation between domain logic and infrastructure, (2) refactoring monolithic or tightly-coupled codebases toward cleaner boundaries, (3) reviewing architecture for dependency rule violations or layer leakage, (4) establishing testing strategies that isolate domain logic, or (5) evaluating whether hexagonal architecture suits a given problem domain.

Conduct structured implementation postmortems to gather feedback on architecture conformance, library friction, and tooling effectiveness. Use when reviewing completed implementations, PRs, or development phases to surface design gaps, boundary violations, and improvement opportunities. Triggers on requests for code review feedback, implementation retrospectives, architecture audits, or library/tooling evaluations.

Pre-implementation design review framework using the df12 Logisphere crew. Stress-tests system designs, RFCs, ADRs, API proposals, data models, and architecture decisions before code gets written. Each expert examines the design through their specialist lens — structural integrity (Pandalump), alternative approaches (Wafflecat), scaling characteristics (Buzzy Bee), contract design (Telefono), failure modes (Doggylump), and long-term viability (Dinolump). Includes a structured pre-mortem and alternatives checkpoint. Use this skill when asked to review a design document, RFC, ADR, system proposal, API design, or architecture decision — or when asked "should we build it this way", "what could go wrong", "design review", "pre-mortem", "architecture review", "RFC review", or any request for pre-implementation feedback.

Community-of-experts review framework using the df12 Logisphere crew for software engineering tasks. Each expert brings a distinct engineering perspective: architecture (Pandalump), creative alternatives (Wafflecat), performance and observability (Buzzy Bee), type safety and contracts (Telefono), reliability and ops (Doggylump), and developer experience (Dinolump). Use this skill when asked to review code, design systems, evaluate architecture decisions, debug complex issues, assess production readiness, or when a thorough multi-perspective engineering analysis is needed. Triggers include: "review this", "what do you think of this design", "is this production-ready", "logisphere review", "expert review", "community review", "crew review", or any request for comprehensive engineering feedback.

Author Open Dynamic Workflows (ODW) workflow scripts. Use when the user wants to design, write, review, repair, or run-check an ODW workflow; when they mention `odw`, Open Dynamic Workflows, Claude Code workflow dialect, dynamic workflows, multi-agent orchestration, `agent()`, `parallel()`, `pipeline()`, nested `workflow()`, workflow generation, JSON-Schema agent outputs, adapter routing, `workspaceMode`, `inplace`, multi-provider handoff, or examples such as fan-out/reduce, deep research, adversarial verification, routing, tournament, or loop-until-dry workflows.

Supervise Open Dynamic Workflows (ODW) runs. Use when the user wants to inspect, monitor, debug, pause, resume, stop, list, follow logs for, retrieve results from, or operate the dashboard/API for an ODW workflow run; when they mention ODW run IDs, `odw status`, `odw logs`, `odw result`, `odw serve`, `events.jsonl`, `status.json`, `result.json`, `error.json`, `worker.log`, run directories, dashboard jobs, SSE streams, Claude Code workflow visibility, `workspaceMode`, `inplace`, multi-provider handoff, or workflow supervision.

Test Open Dynamic Workflows (ODW) workflow scripts effectively. Use when the user wants to write, extend, review, or debug tests for an ODW workflow; when they mention testing `odw` workflows, workflow test suites, mock adapters, schema-satisfying mock agents, parse gates, helper-surface extraction, control-loop simulation, `events.jsonl` assertions, or verifying a workflow before a live run. Complements `odw-authoring` (which covers writing workflows) and `odw-supervision` (which covers operating runs).

Semantic version control — entity-level diffs on top of Git. Use when reviewing code changes, understanding what changed in a commit/PR, diffing files, analysing impact of changes, or when line-level diffs are insufficient. Triggers: sem diff, semantic diff, what changed, entity diff, function diff, impact analysis, entity blame, dependency graph, code change summary, what broke, what was renamed, show changes semantically, 语义差异, 实体级别差异

Use Zod 4 idiomatically in TypeScript projects for schema validation, type inference, serialization/deserialization, JSON Schema interop, and error handling. Trigger whenever the user mentions Zod, schema validation in TypeScript, runtime type checking, form validation, API contract validation, z.object, z.string, z.infer, z.parse, or is writing TypeScript code that involves parsing, validating, or transforming untrusted data. Also trigger when migrating from Zod 3 to Zod 4, or when the user references Zod schemas in the context of OpenAPI, JSON Schema, tRPC, React Hook Form, or similar integrations. This skill covers Zod 4.x (including 4.1 codecs, 4.3 features) and explicitly flags Zod 3 assumptions that no longer hold.

Structured code debugging through hypothesis formation and falsification planning. Use when diagnosing bugs, unexpected behaviour, or system failures where the root cause is unclear. Produces a hypothesis document for execution by another agent rather than performing the investigation directly. Triggers on requests to debug issues, diagnose problems, investigate failures, or create debugging plans.

Write in a compressed, evidence-first voice that is blunt, exact, and conversational without drifting into fluff. Use when the user wants sharper prose, forceful argumentation, deadpan wit, Socratic pressure on weak claims, concise rewrites, opinion columns, debate prep, critiques of bad reasoning, or explanations that should sound authoritative without sounding bloated.

This skill is for interface design — dashboards, admin panels, apps, tools, and interactive products. NOT for marketing design (landing pages, marketing sites, campaigns).

Craft high-precision prompts and edit instructions for Nano Banana image workflows, especially when using the local nanobanana MCP tools for generation, editing, character consistency, or multi-image fusion. Use when the task needs structured prompts, reference-role assignment, layout-heavy image specs, typography-heavy images, iterative edit-first refinement, or reliable model/aspect/output-path choices.

Fast semantic code navigation via LSP. Load FIRST before ANY code task - even 'simple' ones. Trigger scenarios: (1) fixing lint/type/pyright/mypy warnings or errors, (2) fixing reportAny/reportUnknownType/Any type errors, (3) adding type annotations, (4) refactoring or modifying code, (5) finding where a function/class/symbol is defined, (6) finding where a symbol is used/referenced/imported, (7) understanding what a function calls or what calls it, (8) exploring unfamiliar code or understanding architecture, (9) renaming symbols across codebase, (10) finding interface/protocol implementations, (11) ANY task where you'd use ripgrep to find code or read-file to view a function. Use `leta show SYMBOL` instead of read-file, `leta refs SYMBOL` instead of ripgrep for usages, `leta grep PATTERN` instead of ripgrep for definitions, `leta files` instead of list-directory.