Claude Skills by yanacuti1121

How to install and use codebase-memory-mcp — a single-binary MCP server that indexes a codebase into a persistent tree-sitter + Hybrid-LSP knowledge graph (158 languages), answering structural queries (call graph, dead code, HTTP/gRPC/GraphQL route linking, ADRs, Cypher-like queries) in sub-millisecond time with ~10x fewer tokens than file-by-file grep/read exploration. Triggers on: 'index this codebase', 'codebase memory mcp', 'code knowledge graph', 'find callers of', 'call graph', 'dead co...

Verifiable credential lifecycle management for agent identity. Issue, verify, revoke credentials, DID document key rotation, credential schema validation, and W3C VC data model compliance. Sources: microsoft/did-sdk-js, W3C VC spec.

W3C Decentralized Identifier (DID) resolution and cross-agent identity verification. DID document resolution, method routing, key extraction from DID documents, and mutual authentication between Swarm Bus agents. Sources: decentralized-identity/did-resolver.

Disaggregated prefill/decode — separate GPU pools for compute-bound prefill and memory-bound decode. KV cache transfer via NIXL (RDMA/TCP). NVIDIA Dynamo (stack-above) vs llm-d (K8s-native). 6x DeepSeek-R1 on GB200+Dynamo. 30–40% cost savings on $2M+ inference budgets. Sources: rohitg00/ai-engineering-from-scratch (Apache-2.0).

Instrument and debug distributed systems with tracing — OpenTelemetry setup, span creation, context propagation (W3C traceparent), sampling strategy, exporter configuration (Jaeger/Tempo/Honeycomb/Datadog), and correlating traces with logs. Use when asked to "add tracing", "OpenTelemetry", "OTel", "distributed trace", "trace propagation", "traceparent header", "find slow span", "Jaeger", "Tempo", "Honeycomb", "why is this request slow across services", or "trace ID in logs". Do NOT use for: m...

Build production-grade Docker images — multi-stage builds, layer caching, non-root user, minimal base images, .dockerignore, health checks, docker-compose for local dev, BuildKit secrets, and image size audit. Use when asked about "Docker", "Dockerfile", "multi-stage build", "Docker image too large", "non-root Docker", "Docker layer cache", "docker-compose", "Docker health check", "BuildKit", ".dockerignore", "Docker security", "distroless", or "container image best practices". Do NOT use for...

DOMPurify XSS sanitization for HTML content in agent outputs. Strip script tags and event handlers, allowlist-based tag/attribute filtering, sanitize LLM-generated HTML before rendering, and defend against DOM clobbering. Sources: cure53/DOMPurify.

Program LM pipelines with DSPy — define Signatures (input/output fields), build Modules (ChainOfThought, ReAct, Predict), compose them into Programs, then optimize with Teleprompters (BootstrapFewShot, MIPRO, COPRO) to auto-improve prompts from examples.

Durable background job and task queue patterns for AI agent systems. BullMQ Redis-backed queues with concurrency + rate limiting, Inngest event-driven durable functions, Trigger.dev background tasks, dead-letter queue patterns, and job failure diagnostics. Sources: taskforcesh/bullmq, inngest/inngest, triggerdotdev/trigger.dev, OptimalBits/bull, agenda/agenda.

Task-local harness system — declare owns/consumes/produces/pass-fail before dispatch, extract to shared skill when repeated. Use instead of inline ad-hoc steps for any task spanning >1 session or >3 files. Triggers on: 'dynamic workflow', 'task harness', 'harness per task', 'custom harness', 'adaptive workflow', 'task-local harness', 'owns consumes produces', 'harness template', 'dynamic mode', 'workflow harness'.

eBPF-based runtime behavioral monitoring for detecting sandbox escapes and privilege escalation. Tracee event signatures, file access monitoring, network connection alerts, and process execution tracing. Sources: aquasecurity/tracee.

Monitor and intercept Linux kernel syscalls from agent processes using eBPF (Extended Berkeley Packet Filter). Real-time syscall filtering, PID-based policy enforcement, cgroups v2 throttling, and seccomp profile generation.

Security scanner cho Claude Code config — quét CLAUDE.md, settings.json, hooks, MCP servers, agents tìm lỗ hổng bảo mật. 1282 tests, 102 static analysis rules.

Navigation guide cho ECC — onboarding, tìm skills/agents/commands phù hợp, setup, troubleshooting. Dùng khi cần hỏi 'làm X với ECC như thế nào?'

Elliptic curve cryptography for agent key pairs and command signing. EC key generation (secp256k1/p256), ECDSA sign/verify, ECDH shared secret, DER/PEM encoding, and hardware-safe key storage patterns. Sources: indutny/elliptic.

Audit cost overrun cho ECC Tools GitHub App — investigate runaway PR creation, quota bypass, premium-model leakage, duplicate jobs, billing spikes.

Everything Claude Code (ECC) — agent harness configuration layer for Claude Code and other AI coding IDEs. Augments agents with skills, hooks, persistent memory, model routing, and quality gates without modifying the underlying model.

Enterprise-grade component architecture patterns from 20 production design systems. Component naming, data-dense layouts, headless patterns, accessible forms, and dev-tool UI conventions. Sources: IBM Carbon, Adobe React Spectrum, Palantir Blueprint, GitHub Primer, Microsoft Fluent, Salesforce, Workday Canvas, Twilio Paste, Pinterest Gestalt, and 11 others.

Envoy L7 proxy filter chains, xDS dynamic configuration, rate limiting, request mirroring, and Lua/WASM filter patterns for sidecar-level traffic control in agent networks. Sources: envoyproxy/envoy (Apache-2.0).

Systematically remove malware, backdoors, and attacker persistence mechanisms

Production error handling — typed errors, Result type, error boundaries, structured logging, retry strategies

Design structured application error handling — error type hierarchy, error codes, propagation strategy (throw vs return), user-facing vs internal messages, HTTP error response format, async error safety, and React error boundaries. Use when asked to "improve error handling", "structured errors", "error codes", "error types", "don't swallow errors", "unhandled promise rejection", "error boundary", "error response format", "what to show users when something fails", or before shipping any featur...

ESLint custom rule architecture. AST visitor rules with auto-fix, rule metadata, nested rule hierarchies, RuleTester harness, and monorepo-wide static analysis configuration. Sources: eslint/eslint (MIT).

ESTree standard AST node types for JavaScript analysis. Node structure reference, custom AST walkers, node type guards, and building AST-based tools from the official ECMAScript AST specification. Sources: estree/estree (MIT).

etcd distributed key-value store for centralized agent config and distributed locking. Watch API for live config reload, lease-based TTL keys, transactions, and distributed mutex patterns. Sources: etcd-io/etcd (Apache-2.0).

Eval-driven agent development — 3-layer evaluation (static benchmarks, custom offline, online production). Evaluator-optimizer tight loop. Evals in CI, score-gated PRs. SWE-bench/GAIA/BFCL V4 for cross-model comparison. Trajectory-based and LLM-as-judge evals. Sources: rohitg00/ai-engineering-from-scratch (Apache-2.0).

'Evaluates and selects Threat Intelligence Platform (TIP) products based

Design event-driven systems — event sourcing, CQRS, saga pattern for distributed transactions, message queue patterns, and event schema design. Use when asked about "event sourcing", "CQRS", "saga", "distributed transaction", "message queue", "Kafka", "publish/subscribe", "event-driven", or "how to coordinate across services without direct calls". Do NOT use for: real-time WebSocket UI patterns — that is a frontend concern.

Compound Engineering plugin cho Claude Code + Cursor — 37 skills, 51 agents, triết lý 80% planning/review + 20% execution. Mỗi unit làm việc sau dễ hơn trước.

'Executes authorized attack simulations against Active Directory environments

'Executes authorized phishing simulation campaigns to assess an organization''s

Use when the user has approved a plan and says 'go', 'execute', 'proceed', 'do it', or 'run the plan'. Enforces step-by-step execution with no scope expansion. Companion to plan-first skill.

Red team engagement planning is the foundational phase that defines scope,

'Executes comprehensive red team exercises that simulate real-world adversary

Exploit misconfigured Active Directory Certificate Services (AD CS) ESC1

BloodHound is a graph-based Active Directory reconnaissance tool that

'Tests APIs for injection vulnerabilities including SQL injection, NoSQL

'Analyzes and simulates BGP hijacking scenarios in authorized lab environments

'Tests APIs for Broken Function Level Authorization (BFLA) vulnerabilities

Discover and exploit broken link hijacking vulnerabilities by identifying

Exploit Kerberos Constrained Delegation misconfigurations in Active Directory

'Tests and exploits deep link (URL scheme and App Link) vulnerabilities

'Tests APIs for excessive data exposure where endpoints return more data

Detecting and exploiting HTTP request smuggling vulnerabilities caused

Identifying and exploiting Insecure Direct Object Reference vulnerabilities

'Identifies and exploits insecure local data storage vulnerabilities

Identifying and exploiting insecure deserialization vulnerabilities in

'Identifies and exploits IPv6-specific vulnerabilities including SLAAC

'Exploits JWT algorithm confusion vulnerabilities where the server''s

Perform Kerberoasting attacks using Impacket's GetUserSPNs to extract