Agent skills for Perplexity

Migrates Vendure Admin UI extensions (legacy Angular-based) to the new React Dashboard.

Run deterministic financial models for startup valuation and SaaS health analysis.

Access comprehensive LaTeX templates, formatting requirements, and submission guidelines for major scientific publication venues (Nature…

Deploy and manage applications on Vercel, including preview deployments and deployment protection.

Vercel KV (Redis-compatible key-value storage via Upstash).

Use when an agent does structured Pydantic extraction from an LLM and deterministic post-checks are cheap.

Independently verify one ready-for-verification agent ticket against its scope, repository state, real-interface tests, and fresh evidence.

SeldonFrame merge gate — run unit tests + tsc + check-use-server + migration-journal + regression-grep and return ONE pass/fail verdict.

Multi-AI requirements scoping using available external providers (Double Diamond Define phase)

Boolean logic and digital circuits with exact results — truth tables, minimal simplification, Karnaugh maps, normal forms…

Natural-language undo, rollback, and checkpoints that hide git entirely — "ย้อนกลับ", "undo", "ไม่เอาแล้ว", "เอาอันเดิมคืน" map to safe…

Dataset version control for research reproducibility.

Use when a Godot/.NET task needs a Layer 4 overlay upgrade review that evaluates engine, SDK, C# compatibility, addons/plugins, export…

Use when optimizing for Answer Engine Optimization — LLM citation tracking and E-E-A-T.

Use when designing multi-agent orchestration — patterns for delegation and coordination.

Use when applying web best practices — security headers, modern APIs, error handling.

Use when setting up or modifying build and deploy pipelines — shift left, quality gates.

Use when output quality drops or you switch tasks — feed agents the right information at the right time.

Use when diagnosing or optimizing LCP, INP, CLS — Core Web Vitals at the 75th percentile (LCP ≤2.5s, INP ≤200ms, CLS ≤0.1).

Use when implementing changes touching multiple files — thin vertical slices, feature flags, rollback-friendly.

Use when the request is underspecified — interview one question at a time until 95% confidence.