Claude Skills by dayvisonassis

> **Resumo da aula em formato didático.** Este documento traduz os pré-requisitos e o > processo recomendado para construir uma *skill* de agente — aquela que codifica as > decisões arquiteturais do projeto em instruções reutilizáveis pelo agente de IA. ---

Performs a comprehensive surface-area analysis of any codebase scope (a component, module, service, layer, or the full project). Inventories every entry point, dependency, pattern, and integration, scores complexity, and produces a structured Architecture Analysis Report consumed by the deep-analyzer skill. Analysis/reporting only — never modifies the codebase. In the SDD flow this is the first setup step for an existing (brownfield) project; it feeds deep-analyzer and then gate-builder.

Use when the active Claude Code account must change - the work account hit its 5-hour usage limit, the session should continue under a personal account, or a new account must be registered for later switching. Windows, Linux and macOS; only for accounts belonging to the same person.

Answers which files to refactor first to reduce systemic risk, by crossing blast radius (how many modules stop working if this one breaks) against size. On first use it asks permission, installs its own tooling and writes a config by reading the project; on every run after that it executes a versioned script and reads only that script's JSON output, never the source tree again. Produces a standalone HTML report with a scatter chart, a ranked queue and structural detectors. Unlike architecture...

Performs an exhaustive, item-by-item deep analysis of every module, service, component, or endpoint listed in the Architecture Analysis Report (REQUIRED as input). Maps every implementation detail, data flow, contract, and failure mode, and produces a structured Deep Analysis Report. Analysis/reporting only — never modifies the codebase. In the SDD flow it runs after architecture-analyzer and feeds gate-builder.

Externally evaluates an already-implemented feature against its contract.md (environment, quality gates, coverage manifest, observable criteria), producing screenshots, a report, and chat findings. Owns the evaluation loop — keeps the attempt counter in progress.json, decides the state (CLEAN/FAIL/PENDING/ABORTED), and routes each failure: code failures (gate/observable) to fix-runner, test failures to the matching test-writer (then confirmed by the matching test-validator) before resuming. L...

Specialized, lightweight corrector dispatched by the evaluator on FAIL, or by qa-preflight after a feature is done. Reads the structured evaluation-report.json (both producers write the same schema), applies the minimal fix for the reported problem, locally revalidates the failing gate/test, commits (one commit per correction), records the attempt in progress.json, and returns control to whoever dispatched it. Does not implement new features, does not orchestrate the attempt loop, and does no...

Builds the deterministic quality-gate infrastructure for a project (typecheck, lint with zero-warnings, dependency/architecture checks, tests, dead-code, and a single runGate entry point) from the Deep Analysis Report (+ Architecture Analysis Report). Auto-detects greenfield vs. brownfield, proposes a gate plan for confirmation, then writes the configs/scripts and verifies each gate runs. The gates it builds are later declared in each feature's contract.md by spec-writer.

Compacts the current conversation into a handoff document a fresh agent can pick up from, and asks where to save it before writing anything. Use when a session ends with work unfinished, when switching branch or machine, or when passing the work to another person. Records what is not already written down elsewhere and points at what is. Never invoked on its own initiative — a handoff is a deliberate act, and only the person having the conversation knows when it is worth one.

Implements a feature autonomously based on its spec, plan, AND contract.md, committing one commit per phase, satisfying the contract's quality gates and observable criteria, writing progress.json, and reporting results against the feature's acceptance criteria. Delegates test writing to the matching test-writer skill (unit/integration/monorepo/e2e) by path/stack when the project fits, and falls back to writing tests itself otherwise.

Audits PABX backend integration tests (apps/backend/__tests__/integration/) against the integration-test-writer rules, producing a compliance report with a PASS/FAIL/PASS WITH WARNINGS verdict, per-violation severity, and fix suggestions. Read-only — never writes or fixes tests. Dispatched by the evaluator to confirm a corrected integration test conforms before resuming evaluation, or run directly by a user.

Plans and implements comprehensive integration tests for the PABX backend (apps/backend/__tests__/integration/) using Jest + supertest against a real test database. Runs interactively (3 phases with an approval checkpoint) when a user calls directly, or autonomously (no pause) when dispatched by implement-feature/evaluator. Also runs in correction mode to fix a specific failing integration test flagged by the evaluator. Enforces PABX rules (setupTestDatabase, try-finally cleanup, foreign-key ...

Audits PABX monorepo unit tests (node-express, node-worker, python-fastapi apps outside apps/backend and apps/frontend) against the monorepo-unit-test-writer rules, auto-detecting the stack and producing a compliance report with a PASS/FAIL/PASS WITH WARNINGS verdict, per-violation severity, and fix suggestions. Read-only — never writes or fixes tests. Dispatched by the evaluator to confirm a corrected test conforms before resuming evaluation, or run directly by a user.

Plans and implements unit tests for any PABX monorepo app EXCEPT apps/backend and apps/frontend — auto-detecting the stack (node-express, node-worker, or python-fastapi) and applying stack-specific patterns. Runs interactively (3 phases with an approval checkpoint) when a user calls directly, or autonomously (no pause) when dispatched by implement-feature/evaluator. Also runs in correction mode to fix a specific failing test flagged by the evaluator. Enforces PABX rules (mock-before-import, P...

Fonte-da-verdade do design system do DRCALL PABX (Angular Material M3, aesthetic compacto/profissional). Use ao criar ou revisar qualquer UI do frontend — páginas, tabelas, dashboards, formulários, diálogos — para seguir tipografia, espaçamento, tokens de cor, padrões de página/tabela/dashboard e os anti-padrões proibidos. É também a base do gate styles-frontend e da verificação visual.

Generate complete PRDs (Product Requirements Documents) through iterative clarification. Use when: (1) Starting a new project and need structured requirements, (2) Creating product specifications with 9-section PRD format, (3) Defining product scope from a description or context files. Keywords: "prd", "product requirements", "create PRD", "generate PRD", "new product", "requirements document".

The gate between a finished feature and the human QA. Investigates the delivered feature against its spec and the running screen, executes what is automatable, dispatches the existing correctors for defects that are objectively wrong, leaves a permanent guard so each defect cannot return unnoticed, and writes a QA plan (md + csv) plus a findings report carrying only what needs human judgement. Invoked by a HUMAN when the feature is done — never by spec-writer, implement-feature or evaluator.

Generates technical implementation spec, plan, AND an operational contract for one or more features based on PRD, codebase analysis, and iterative clarification. Produces contract.md (environment contract, quality gates, coverage manifest, observable criteria) consumed by implement-feature and evaluator. Supports batch mode for generating multiple features from the same wave in parallel.

Audits unit tests in the PABX monorepo (Angular frontend .spec.ts + Node.js backend __tests__/unit/) against the unit-test-writer rules, producing a compliance report with a PASS/FAIL/PASS WITH WARNINGS verdict, per-violation severity, and fix suggestions. Read-only — never writes or fixes tests. Dispatched by the evaluator to confirm a corrected test conforms before resuming evaluation, or run directly by a user.

Plans and implements comprehensive unit tests for the PABX monorepo — Angular frontend (Jest + jest-preset-angular, .spec.ts) and Node.js backend (Jest, apps/backend/__tests__/unit/). Runs interactively (3 phases with an approval checkpoint) when called by a user, or autonomously (no pause) when dispatched by implement-feature/evaluator. Also runs in correction mode to fix a specific failing test flagged by the evaluator. Enforces PABX rules (real component imports, manual service instantiati...

Audits PABX end-to-end tests (tests/e2e/) against the e2e-test-writer rules, producing a compliance report with a PASS/FAIL/PASS WITH WARNINGS verdict, per-violation severity, and fix suggestions. Read-only — never writes, fixes or runs tests. Dispatched by the evaluator or qa-preflight to confirm that an e2e test a writer corrected or newly wrote (a missing-coverage test, a guard) conforms before they trust it, or run directly by a user.

Plans and implements end-to-end tests for the PABX monorepo (tests/e2e/) using @playwright/test against the running app — one user flow per test, an action and its observable result, traced to the contract's surfaces and criteria by tags. Runs interactively (3 phases with an approval checkpoint) when a user calls directly, or autonomously (no pause) when dispatched by implement-feature/evaluator/qa-preflight. Also runs in correction mode to fix a specific failing e2e test, and in guard mode t...