mcampbellr/engineering-agents
Twelve engineering review and audit skills with shared evidence reporting and domain-specific workflows.
Audit responsive web, mobile and desktop layouts for overflow, breakpoints, safe areas, keyboard visibility, reflow and window resizing.
Review LLM integrations, prompts, token budgets, agent orchestration, observability, evaluations and AI security. Use for AI/LLM usage audits.
Find unintended implementation outliers among semantically comparable files. Use for codebase consistency audits; report before optional remediation.
Audit data ownership, schemas, invariants, module boundaries and concurrency correctness. Use for data architecture and code organization reviews, not stylistic folder rearrangement.
Review infrastructure sizing, unused resources and operational cost risks. Use for infrastructure cost audits; LLM token costs belong to ai-audit.
Audit memory retention, bounded concurrency, streaming, pagination and cache correctness for large datasets. Use for memory and bulk data reviews; distinguish measured leaks from static risks.
Audit web/native client loading, rendering, bundles, requests and runtime performance. Use for client slowness; backend capacity belongs to scalability-audit.
Review changes for duplicated implementations and missed compatible reuse, with evidence for both locations. Use for DRY and reuse reviews.
Audit system scalability, throughput, tail latency, overload and resilience using workload evidence. Use for architecture or backend capacity reviews, not a standalone client rendering audit.
Audit sensitive files, secrets, dependency vulnerabilities, supply-chain integrity and CI/CD security. Use for repository security scanning; application authorization belongs to web-security-audit.
Review a scoped diff through three independent static lanes, deduplicate findings and apply the existing remediation policy. Use when Superreview is explicitly requested.
Audit web application attack paths, authentication, authorization, business logic and tenant isolation with OWASP verification evidence. Use for defensive application security reviews; not a secret, dependency or CI/CD scanner.