The library

claude-mem · Handling specialized tasks that generic agents cannot perform.

firefox-devtools · Handling specialized tasks that generic agents cannot perform.

iwdp-mcp · Handling specialized tasks that generic agents cannot perform.

pagemap · Browsing high-token pages (PDFs, documentation, dense paywalls) where 97% compression outweighs loss of formatting.

cloudbase-ai-toolkit · Tencent-native development where CloudBase platform provides cost-effective Alibaba-competitive serverless infrastructure for China-market apps.

cloudbase-sites · Two-stage save→deploy workflow for iterative React/Vite front-end development on CloudBase, inspired by Codex Sites pattern.

cloudbase · Tencent stack full-stack apps needing unified auth, NoSQL+SQL, blob storage, serverless, and WeChat Mini Program integration.

forge · Multi-file refactors, full-stack features, or spec-driven work where parallel validation and retry beat sequential manual oversight.

desktop-commander · Local-only automation across text, code, PDFs, spreadsheets where client-side execution avoids cloud egress and latency.

pty-mcp · Long-lived interactive debugging, DevOps orchestration, or system administration where persistent PTY state beats stateless shell commands.

vpe-advisor · VP of Engineering or Director of Eng optimizing delivery throughput (DORA), hiring funnel conversion, and org structure when execution velocity stalls.

engineering-skills · When shipping production-grade code requires discipline across architecture, testing, deployment, and security domains rather than isolated linting.

a11y-audit · When shipping accessible UI requires automated detection of common WCAG violations before manual testing.

pw · When shipping e2e tests across multiple browsers and platforms requires templates, flaky-failure fixes, and reporter integrations rather than starting from scratch.

snowflake-development · When building production data pipelines in Snowflake requires serverless compute, AI-native SQL functions, and dbt workflow management.

engineering-advanced-skills · When shipping complex agent systems requires principled design across agent logic, retrieval, data layers, operations, and pre-production compliance gates.

agenthub · When a problem benefits from diverse solution approaches and time permits parallel exploration with an objective win criterion.

autoresearch-agent · When optimizing a single file's performance, quality, or cost over time requires hands-off autonomous loops with clear metric tracking.

chaos-engineering · When improving system resilience requires controlled failure injection with documented hypothesis and error-budget impact, not guesswork.

code-tour · When onboarding developers or explaining architecture requires guided walkthroughs tailored to reader expertise level rather than generic comments.

docker-development · When production containers require optimized build time, minimal image size, and security hardening rather than default Docker patterns.

feature-flags-architect · When shipping features requires disciplined flag management with rollout rings, stale-flag cleanup, and audit trails rather than ad-hoc toggles.

grill-with-docs · Crystallizing terminology and architectural decisions incrementally as a plan develops, preserving them in version-controlled docs instead of tribal knowledge

helm-chart-builder · Accelerating Helm chart authoring by enforcing dependency management, security patterns, and testing discipline from the start

karpathy-coder · Catching unsafe abstractions and unfocused changes before code review by surfacing assumptions, complexity drift, and vague goals

kubernetes-operator · Building production Operators by enforcing the Operator pattern disciplines (CRD design, reconcile loops, idempotency, error handling) rather than generic Kubernetes knowledge

prompt-governance · Managing prompt changes in production by preventing silent quality regressions and running structured experiments before rollout

security-guidance · Preventing command injection, XSS, SQL injection, unsafe deserialization, code injection, and GitHub Actions workflow injection at the moment of writing, before code reaches review

slo-architect · Structuring service reliability goals with defensible error budgets instead of ad-hoc alerting or targets disconnected from business risk

terraform-patterns · Structuring Terraform projects with reproducible module layouts, multi-environment state management, and policy enforcement instead of ad-hoc scripts

universal-scraping-architect · Scraping web content with automatic fallback to local extraction when Firecrawl quota is low, avoiding token blowouts on large batch jobs

workflow-builder · Structuring multi-agent work by proposing proven orchestration topologies upfront instead of requiring users to architect their own workflows

write-a-skill · Creating high-quality, reusable skills with consistent structure and documentation instead of ad-hoc prompt files or private scripts

markdown-html-skills · Publishing markdown documents (code reviews, long-form guides, slide decks, presentations) as shareable, visually polished single-file HTML instead of raw markdown or PDF

apple-hig-expert · iOS/macOS teams shipping premium apps when HIG compliance + accessibility + 2026 Liquid Glass aesthetic are critical brand signals.

budgetclaw · Teams running Claude Code agents on metered LLM budgets who need hard spend guardrails with immediate notification of overages.

claude-code-audit-stack · Cloud automation teams that deploy daemons via Claude Code and need adversarial verification that restarts actually succeeded (not just reported success).

claude-hud · Claude Code operators who want immediate visual feedback on token consumption and tool execution without context-switching to inspect dialogs.

claude-snapshot · Teams that onboard developers frequently and want to replicate a standardized Claude Code config across machines without manual setup.

codex-hud · Teams using OpenAI Codex who need quick inline visibility to daily/weekly/monthly costs without switching to OpenAI dashboard.

encode-toolkit · Genomics researchers who need to search ENCODE experiments, download files reliably, and cross-correlate with disease/population databases in a single workflow.

erne-universal · React Native teams that want 8 parallel specialized agents handling architecture decisions, code generation, testing, and deployment without manual orchestration.

give-claude-eyes · Developers who want to add native video/image understanding to Claude Code workflows without calling external vision APIs directly (Qwen abstracts the complexity).

hooks-safety · Teams using Claude Code with destructive git and shell operations where accidental force-push or rm -rf would be catastrophic (startups, financial systems).

kegg-mcp-server · Bioinformatics researchers who need quick access to KEGG's pathway, gene, and disease databases without maintaining local KEGG files or parsing KGML manually.

magic-codex · Large codebases where parallel agent work (e.g., feature team A + feature team B + reviewer) needs isolated environments and safe resumption after interruption.

msapps-cowork-session-fixer · Unblocking a Cowork session failure without manual process restart or support ticket.

msapps-fix-chrome-connection · Quickly restoring Chrome MCP connectivity without uninstalling the extension.

msapps-gcloud-health-check · Pre-flight check before deploying to Cloud Run or debugging GCP permission issues.

msapps-github-health-check · Catching authentication or rate-limit issues before running batch GitHub operations.