voku

Architecture-focused review lens for assessing coupling, module boundaries, abstraction quality, and rollback-safe design during code review. Use when reviewing design quality, separation of concerns, transaction boundaries, or structural maintainability. Triggers on "review architecture", "architecture review", "module boundaries", "transaction boundary", and related review requests.

Error-handling and resilience review lens for catching swallowed failures, unbounded I/O, missing timeouts, retry hazards, and partial-failure bugs during code review. Triggers on "error handling review", "resilience review", "retry logic", "timeout discipline", and related review requests.

Performance-focused review lens for identifying inefficiencies, high-cost query shapes, resource waste, and worst-case scalability issues during code review. Triggers on "performance review", "find bottlenecks", "n+1 review", "resource usage", and related review requests.

Security-focused review lens for identifying injection risks, auth flaws, secret exposure, unsafe configuration, and missing validation during code review. Triggers on "security review", "check for vulnerabilities", "auth review", "injection risks", and related review requests.

Simplicity-focused review lens for identifying unnecessary complexity, dead logic, unclear naming, duplicated behavior, and maintainability drag during code review. Triggers on "simplicity review", "reduce complexity", "readability review", "maintainability review", and related review requests.

Type-safety review lens for catching schema mismatches, unsafe coercions, missing annotations, and weak contracts during code review. Triggers on "type safety review", "review typings", "schema mismatch", "unsafe cast", and related review requests.

End-to-end testing with Playwright for web applications. Use when writing E2E tests, browser automation, form submission testing, or user flow testing. Triggers on "playwright", "e2e test", "browser test", "end-to-end", "form flow testing", or test files in tests/e2e/.

Repo-owned operational prompting for coding agents. Use when designing AGENTS.md, Copilot/Codex instructions, path-specific instruction files, portable skill manifests, or validation contracts. Triggers on "operational prompting", "agent instructions", "repo-owned prompts", "Copilot instructions", or "portable agent skills".

Apply Amy's Echo Amy-first delivery workflow for implementation and review tasks. Use when changing code, tests, or docs in this repository. Enforce discovery, planning, Amy-impact validation, and mandatory verification before merge.

Execute Amy's Echo model training workflows end-to-end with reproducible evidence. Use when changing trainer code, training metadata contracts, model artifact handling, few-shot evaluation, or release-readiness checks for training outputs.

Build and harden Amy's Echo backend APIs using Express 5 and Zod contracts. Use when adding or modifying server routes, middleware, request validation, auth/profile checks, error handling, and response-shape guarantees.

Protect Amy's Echo account-versus-profile identity boundaries. Use when working on auth, settings, profile switching, uploads, training sync, or API flows where session/token and profileId can diverge.

Implement and debug MediaPipe Tasks Vision gesture recognition in Amy's Echo webapp. Use when changing webcam capture, landmark extraction, gesture recognizer setup, model loading, runtime confidence logic, or detection diagnostics tied to MediaPipe behavior.

Diagnose and improve Amy's Echo web sign-language ML detection quality. Use when working on MediaPipe landmarks, MLP inference thresholds, confidence/margin behavior, detector telemetry, fixture diagnostics, or profile-aware recognition regressions.

Run Amy's Echo testing and quality gates with clear evidence reporting. Use when preparing a change for commit/PR, diagnosing regressions, or deciding the minimum safe check set for touched webapp, server, or integration areas.

Configure and enforce Vitest coverage quality gates for Amy's Echo webapp changes. Use when adding tests, tuning coverage thresholds, fixing flaky coverage regressions, or defining test include/exclude behavior in Vite/Vitest.

Compile and use agent-loop recall/L2 meta-prompt artifacts from the current repository without mistaking generated context for executed agent actions.

Use the governed agent-loop workflow for this repository, including when to prefer workflow start/status/close, when to drop to session or recall commands, and how init validation fits repo-managed agent guidance changes.

Start a governed agent-loop task in the current repository, create session working memory, approve a sealed work brief, and compile deterministic recall/L2 context.

Handle reusable knowledge that surfaces during a task — capture findings, move them through the proposal pipeline, and respect the boundary between workflow evidence and durable guidance.

Review, verify, and close an agent-loop task safely after implementation, including blind-spot review, strict verification, and explicit accepted-risk handling.

Record useful working memory during an agent-loop task, including decisions, checkpoints, validation results, scope changes, and blocked states.

Maintain repo-managed agent skills, shared agent docs, metadata roots, and migration-safe validation workflows around agent-loop init.

Capture reusable lessons about agent-loop workflow, validation, and migration boundaries before promoting them into durable agent guidance.

Use ctx-backed local agent-history search as bounded evidence references while creating, validating, or consolidating voku/agent-learning findings.

Set up and operate voku/agent-learning in a consuming repository, including learning-root config.json, findings, proposals, consolidation prompts, validation, and lifecycle commands. Use for end-user learning-loop workflows, not package source development.

Maintain voku/agent-learning package source, tests, docs, CLI behavior, configurable paths, constraint lifecycle code, and local vendor syncs in consuming projects.

Maintain repository-owned agent skills, memory, docs, and installed copies after reviewed durable learnings.

Capture reviewable agent-learning findings for this repository after meaningful sessions.

Turn stable repository guidance into local executable constraints such as static-analysis rules, tests, or CI checks.

Turn validated agent-learning findings into human-approved executable constraints such as PHPStan rules, PHP-CS-Fixer rules, tests, or CI checks, with generation packages, manifests, fixtures, and validation evidence.

Capture, validate, and transition package-specific findings and proposals for voku/agent-learning.

Maintain voku/agent-recall-compiler package source, tests, docs, prompt output, manifest loading, scope selection, outcome logging, and local vendor syncs in consuming projects.

Use voku/agent-recall-compiler in a consuming repository to compile L2 task briefings, select active guidance and constraints by scope, produce validation plans, and log outcomes after a session.

Compile task-scoped recall before coding and log selected/helpful/irrelevant/harmful outcomes afterward.

Capture durable repo-local memory after meaningful discovery, debugging, refactoring, or implementation work.

Capture durable repo-local memory after meaningful discovery, debugging, refactoring, or implementation work.