Explorar skills

Configure email authentication so legitimate mail lands and spoofed mail is blocked. Covers SPF, DKIM, DMARC (with the p=none → p=quarantine → p=reject migration path), MTA-STS, TLS-RPT, ARC, and BIMI. Invoke when launching a new sending domain, when domains are being spoofed, or when transactional email is landing in spam.

Semantic search for finding code by meaning using natural language queries. Orchestrates semantic-search-reader (search/find-similar/list-projects) and semantic-search-indexer (index/reindex/status) agents. Use for understanding unfamiliar codebases, finding similar implementations, or locating functionality by description rather than exact keywords. (project)

Audit and defend against malicious dependencies in npm, pnpm, PyPI, and similar ecosystems. Covers lockfile hygiene, the limits of npm audit, behavior-level scanning with socket.dev, postinstall script review, typosquat and slopsquat detection, and minimum-permission CI runs. Invoke when adding a new dependency, after a supply-chain incident, or as periodic audit.

Apply operational controls to applications built on the Anthropic API or similar LLM SDKs. Maps the OWASP LLM Top 10 to practical controls, plus rate limiting, cost caps, PII scrubbing, audit logging, model-version pinning, and an AI-incident response playbook. Invoke when shipping an LLM feature to production, when handling an abuse complaint, or after a model-provider advisory.

Use when writing or refactoring modern C# with maintainability and performance standards. Keywords: coding standards, C# best practices, refactor C#, records, pattern matching, async best practices.

Use when the user wants text to sound more human, says writing sounds "too AI" or "too ChatGPT," asks to humanize or rewrite a draft to feel natural, or shares content wanting it to feel authentic and less robotic. Also applies to LinkedIn posts, blog drafts, or emails where the user wants a more genuine voice.

Use when you need CRAP score and coverage risk analysis to find high-risk code before refactoring or release. Keywords: CRAP score, risk hotspots, cyclomatic complexity, test coverage gaps, coverage threshold.

Design logging that supports investigations without becoming a privacy liability. Covers what to log and what never to log (PII, secrets), structured logging, retention tiers, centralization choices, alert routing, and the operational-versus-access-versus-audit log split. Invoke when starting a new service, when investigation revealed missing log fields, or when log volume is becoming expensive.

Find Next.js-specific security issues across App Router, Pages Router, and Server Actions. Covers the middleware-bypass class, NEXT_PUBLIC environment leakage, RSC over-fetch, CSP for App Router, open redirects, and next/image SSRF via permissive remotePatterns. Invoke when reviewing a Next.js app before launch, after a major version upgrade, or when adding authenticated routes.

Audit Model Context Protocol server configurations and apply least-privilege scoping. Covers MCP inventory, capability risk-tiering, secret detection in configuration, malicious or compromised package indicators, and the lifecycle from install through rotation to revocation. Invoke before granting an MCP write access to production, after an MCP security advisory, or as periodic audit.

Use when reviewing PRs/diffs/branches/documents for .NET quality, correctness, performance, security, data access, messaging, and observability. Includes adversarial critical-thinking mode for skeptical/cynical review requests. Keywords: code review, review PR, review diff, critical review, cynical review, adversarial review, production readiness, low GC, AOT, security review.

Design backends that survive redeploys, server reboots, and modest scaling. Covers stateless application servers, state placement (object storage, managed databases, Redis), immutable deploy artifacts, health checks, graceful shutdown, database migrations that don't lock the world, and the twelve-factor baseline. Invoke when designing a new backend, when uploads disappear after a redeploy, or when