Development

Layer A interaction-mechanics reference anchored to the beui.dev catalog. Stacks on any style skill whenever work adds or changes motion or interaction — micro-interactions, animated components, transitions, gestures, hover/press/state feedback, loading/success/error morphs, 'make it feel alive'. Mandates reading the mapped beui.dev component source before designing an interaction; owns interaction mechanics and reduced-motion discipline; owns zero visual taste. Load it alongside a style skil...

Publish a finished lesson MP4 from ~/Movies/lpm-lessons/<slug>/ to the lpm YouTube channel through Chrome (YouTube Studio): search-optimized title and description with chapters, tags, category, the title-card thumbnail, Public, then added to the lesson playlist. Use when the user asks to upload, publish or post a video to YouTube.

Make a narrated lesson video about lpm from the real desktop app, recorded on a pristine data directory. Title and optional narration text in, MP4 with OpenAI voice out, audio in sync with the clicks. Use when the user asks for a video lesson, tutorial, screencast, or YouTube video about lpm.

Audit a project's Claude Code configuration (.claude/ directory, CLAUDE.md/AGENTS.md, settings.json, MCP server configs, hooks, and agent definitions) for security vulnerabilities and misconfigurations — hardcoded secrets, overly permissive tool allow lists, command injection in hooks, prompt-injection surface. Use whenever a hooks.json, agent roster, or settings.json changes, and before publishing a plugin/engine publicly. Newly relevant now that this engine ships real hooks and an agent ros...

Use before claiming a feature, refactor, or fix is complete, and before opening a PR — a fixed six-phase mechanical gate (build, types, lint, tests, security, diff review) that produces a READY/NOT READY verdict. Not a test-design framework — use `sdlc-meta/advanced-testing-strategy` to decide what tests to write; this is the ordered pipeline that runs them and reports the result.

Use when a defect, regression, intermittent failure, or performance symptom needs evidence-led reproduction and causal diagnosis before a fix; use advanced-testing-strategy to design broader test coverage.

Use when a long session is approaching a context limit, or is between task phases (research/plan/implement/test), to decide whether this is a good point to manually compact — rather than waiting for arbitrary auto-compaction to hit mid-task and discard state you actually needed.

Use when output must clear two independent adversarial reviewers before it ships — production code, published content, or any deliverable where a single self-reviewing agent shares the blind spots that produced the output. Not for internal drafts or deterministic checks (use build/test/lint pipelines for those).

Decision framework for choosing between regex and an LLM call when parsing structured or semi-structured text (forms, invoices, quiz/exam content, scraped listings) — start with regex, add an LLM only for the low-confidence remainder. Relevant to this engine's scraping and document-processing work.

Use when a task can go faster by doing independent work at the same time — batched reads/searches, concurrent subagents, isolated worktrees, or multiple verification lanes — without letting concurrency create conflicting writes or a false sense of "done." Not for deciding whether a task should be delegated to a subagent at all — this is about running already-independent lanes efficiently once decomposed.

Fork, sanitize, and package a private project for safe public release — a 3-stage pipeline (fork/strip secrets, sanitize/verify clean, package for release). Use when a private project (including any of this consultancy's engines or client codebases) must be prepared for public GitHub release. Prerequisite before publishing any Chwezi engine publicly.

Use when onboarding an AI coding agent onto a hand-written legacy project (including this consultancy's WAMP/PHP client codebases) and you need to prevent "style drift" — the model imposing pretrained mainstream idioms onto a project with its own implicit conventions. Language- and framework-agnostic; aligns meta-architecture only, not syntax. Once run, becomes a standing behavioral constraint on all subsequent coding tasks in that project. Not for pure research or one-off questions unrelated...

Use for GitHub repository operations beyond plain git — issue triage, PR review status and stale-PR policy, CI/CD failure diagnosis, release and changelog management, and Dependabot/secret-scanning monitoring, all via the `gh` CLI. Not for branching strategy, merge-vs-rebase, or local conflict resolution — use `sdlc-meta/git-collaboration-workflow` for those.

Use before answering setup, configuration, or API-reference questions about a specific library or framework, or before writing code that depends on its exact current behavior — fetch live docs via an MCP documentation server (e.g. Context7) instead of relying on training data, which is a fixed snapshot and goes stale as libraries release new major versions.

Convene a four-voice council for ambiguous decisions, tradeoffs, and go/no-go calls where multiple valid paths exist and structured disagreement is more useful than a single confident answer. Not for code review, implementation planning, or architecture design — use the dedicated agents/skills for those.

Create CodeTour `.tour` files — persona-targeted, step-by-step walkthroughs with real file and line anchors. Use for onboarding tours, architecture walkthroughs, PR tours, RCA tours, and structured "explain how this works" requests.

Use when architecting, specifying, implementing, or auditing PMS, POS, restaurant, hotel, resort, lodge, guest-house, venue, catering, or food-service software.

Use when building, diagnosing, or modernising Java/JVM APIs, data pipelines, modular ERP, or Hibernate N+1/fetch plans with Spring Boot/Batch, Jakarta EE, transactions, tests, PostgreSQL/jOOQ, Maven/Gradle, WebLogic, Oracle JDBC, Quarkus, GraalVM, JFR, or javax-to-jakarta.persistence migration.

Design, implement, and refactor Ports & Adapters systems with clear domain boundaries, dependency inversion, and testable use-case orchestration across TypeScript, PHP, Java, Kotlin, and Go services. Use when introducing or refactoring toward Ports and Adapters, or when domain logic has become entangled with I/O. Complements skills/architecture/microservices-architecture (service topology) and skills/architecture/api-design-first (boundary contracts) — this skill governs internal module bound...

Progressive multi-pass context refinement for a context-starved subagent — a 4-phase DISPATCH-EVALUATE-REFINE-LOOP cycle, max 3 cycles, for when a subagent cannot predict upfront which files or context it needs. Narrower than skills/ai/ai-rag-patterns (general agentic RAG) and skills/ai/ai-agent-multi-agent-coordination (handoff contracts) — this skill is specifically the retrieval-refinement loop inside one subagent's own context-gathering pass.

Dense financial-newspaper theme with a compact serif masthead, restrained lead-story grid, ruled story columns, numbered most-read coverage, and a continuous light-grey newsprint surface.

A Claude-powered CI review check reports red. Use when triaging it, to separate a genuine finding the check could not publish from turn-budget exhaustion — read is_error first, never the red X.

Decode any Base transaction into a plain-English story - method, token movements, swaps/approvals, counterparties, and suspicious-approval flags. Keyless via Base RPC + Etherscan v2.

Daily on-page and technical SEO audit of every page on a site - discovers URLs from the sitemap, scores each page, adds cross-page checks (duplicate titles, canonicals, sitemap gaps), diffs against yesterday, and sends the score line plus any regressions