fund enables wallet top-ups via coinbase onramp with multiple payment methods (apple pay, card, bank transfer, coinbase account) or direct usdc transfers on base network.

firebase-storage covers client and server file operations including uploads with progress tracking, resumable transfers, downloads, signed URLs, and security rule templates for access control patterns.

ml-pipeline-automation covers end-to-end orchestration of data ingestion, processing, training, validation, and deployment using airflow with mlflow experiment tracking and modular python functions.

amap-lbs-skill wraps six map service scenarios (poi search, nearby search, heatmap visualization, detailed poi lookup, route planning, travel planning) with url generation and api integration patterns. execution relies on external scripts and keys without robust error handling.

image-inpainting enables mask-driven region edits (object removal, background replacement, retouching) via runcomfy cli, routing to z-image turbo when masks are provided and falling back to description-based models otherwise.

platform-sharing-rules-generate handles creation, modification, and deletion of salesforce sharing rules metadata across criteria-based, role-based, and guest user sharing types.

gitlab-ci-patterns provides yaml templates for multi-stage pipelines, docker builds, kubernetes deployments, and terraform workflows. examples are functional but lack step-by-step guidance and failure recovery patterns.

ai-avatar-video routes audio and images across five runcomfy models (omnihuman, wan, happyhorse, seedance) to generate talking-head and avatar videos, selecting the appropriate model based on user intent like dubbing, ugc ads, or character animation.

vue-best-practices encodes composition api and typescript conventions for vue 3 development. establishes architecture decisions upfront and references core patterns, but execution guidance remains shallow and incomplete.

stream-chain orchestrates sequential multi-agent workflows where each step receives prior output as context. supports custom prompt chains and four predefined pipelines (analysis, refactor, test, optimize) with configurable timeouts and verbosity.

m10-performance guides measurement-first performance work, mapping profiling results to specific optimization techniques (algorithm, data structure, allocation, cache) with estimated impact ranges and anti-pattern warnings.

agent-tool-builder addresses schema design and error handling for ai agent tools, emphasizing clear descriptions and validation patterns to prevent hallucination and silent failures.

llm-app-patterns provides code-heavy reference material for rag pipelines, agent architectures, and prompt management. covers multiple approaches but lacks actionable decision frameworks and explicit failure handling.

claude-code-guide outlines configuration templates and prompting patterns for steering claude code agent behavior, but lacks concrete decision logic and failure recovery steps.

pytest skill provides broad tutorial coverage of python test framework patterns including fixtures, parametrization, markers, and integration with fastapi/django. primarily educational reference material rather than an executable task.

chinese-code-review provides templates and tiered feedback categories (must-fix, suggest, reference-only) for culturally-aware code review in mainland teams, explicitly triggered via /chinese-code-review command only.

site-audit provides webflow site analysis covering pages, cms collections, and health metrics, but lacks explicit decision logic and recovery paths for common failure modes.

stripe-payments guides integration of stripe payment flows into web apps, covering checkout sessions, payment intents, subscriptions, and customer management with decision framework and working code examples.

design-system-patterns covers token hierarchies, theming infrastructure, and component architecture for building consistent ui foundations. emphasizes naming conventions, multi-platform generation, and figma-to-code workflows but lacks actionable step-by-step procedures.

copy-editing applies a seven-pass framework to refine marketing text without rewriting it. the skill reads product context files when available and works through clarity, voice, benefits, proof, specificity, emotional tone, and risk systematically.

building-sf-integrations covers salesforce integration architecture including named credentials, external services, callout patterns, and event-driven design, with clear ownership boundaries but minimal procedural depth.

data-pipeline documents etl workflow patterns using n8n, covering extraction from apis and databases, transformation logic, and loading to warehouses. includes scheduling, monitoring, and data quality checks but lacks concrete decision trees and failure recovery procedures.

boost-prompt guides iterative refinement of task prompts through structured questioning around scope, deliverables, and constraints, outputting polished markdown to clipboard via vscode integration.

grill-with-docs runs relentless design interrogation that stress-tests plans against domain models, grounds discussions in codebase artifacts, and produces focused documentation (ADRs, glossary updates) only when decisions crystallize.

git-pushing stages, commits with conventional format, and pushes changes to remote. triggers on explicit push requests, save-to-remote mentions, or feature completion signals.

remotion-upgrade handles version detection and conditional package updates across remotion, mediabunny, and related skills, with fallback to manual upgrade when cli is unavailable.

dingtalk-ai-table covers crud operations on dingtalk's notable api multi-dimensional tables, including sheet/field/record management. delegates actual api formats to external references and requires manual token/operatorid provisioning.

sentry-setup-ai-monitoring auto-configures monitoring for llm calls across python and javascript stacks. detects installed ai packages (openai, anthropic, langchain, vercel ai) and enables tracing with zero manual registration. captures model, tokens, and latency by default with opt-in pii controls.

llm-wiki describes a three-layer knowledge distillation pattern for maintaining an obsidian vault via llm synthesis. the architecture separates immutable sources, synthesized wiki pages, and query interfaces, but the skill lacks concrete procedural steps and triggering conditions.

orient fetches github issue context and auto-detects task type from labels to map a git branch prefix, accelerating task initialization and branch naming consistency.

flutter-embedding-native-views covers platform view integration across android, ios, and macos, plus web embedding modes. includes composition tradeoffs, step-by-step workflows, and performance mitigation strategies.

platform-custom-object-generate covers salesforce custom object metadata creation, xml generation, and field/validation rule management with required description freshness enforcement on every object mutation.

ziniao-device manages proxy devices (http, https, socks5, ssh, ssl) for the ziniao browser. covers querying, adding, modifying, binding to stores, and handling purchase/renewal/auto-renewal with high-risk write confirmations.

agent-memory provides file-based persistent storage for cross-conversation knowledge using markdown files with frontmatter. covers save, search, maintain workflows with ripgrep queries and folder organization patterns.

figma-generate-design guides composition of screens and multi-section layouts from existing design systems, emphasizing component reuse over primitives. requires figma-use as a prerequisite and defines clear skill boundaries against component creation and code mapping tasks.

security covers attack surface mapping and vulnerability fixes in dune apps, with explicit focus on auth gates, external data entry points, and credential handling, though execution guidance remains incomplete.

upstash-workflow describes three core async patterns (dry-run mode, fan-out, single task execution) for qstash integration in lobehub, emphasizing rate limit and idempotency constraints but lacks concrete implementation steps.

context-network provides high-level framing for organizing project knowledge across sessions, but lacks concrete procedural steps. covers diagnosis and scaffolding generation without detailing how execution happens.

design-critique provides a structured framework for evaluating mockups across usability, hierarchy, consistency, and accessibility. strong template and clear critique dimensions, but execution depends entirely on user input quality and the skill lacks procedural guardrails.

statistical-hypothesis-testing provides python implementations of parametric and non-parametric tests (t-test, anova, chi-square, mann-whitney) with effect size calculations, confidence intervals, and power analysis. covers test selection, assumptions validation, and multiple comparison corrections.

extract identifies repeated ui patterns, hard-coded values, and design tokens scattered across a codebase, then consolidates them into a cohesive design system with incremental adoption guards and accessibility safeguards.

wordpress-performance-best-practices collects 34 optimization rules across database, caching, and asset management for plugin and theme development. references correct/incorrect code patterns but lacks prescriptive step-by-step workflows.

rhwp-edit modifies hwp document text, tables, and cells via k-skill-rhwp cli wrapping a rust/wasm engine, preserving formatting across insertText, deleteText, replaceAll, createTable, and setCellText operations.

crypto-market-rank exposes five leaderboard endpoints (social hype, token rank, smart money inflow, meme tokens, address pnl) through a unified cli that maps subcommands to upstream rank aggregators. coverage across multiple ranking dimensions but thin on execution details.

firecrawl-download maps and scrapes entire websites into organized local file trees, supporting markdown, screenshots, and multi-format output per page with path filtering and format control.

roadmap generates multi-phase delivery blueprints for application builds, suitable for autonomous claude code execution across sessions. each phase contains concrete, ordered tasks designed to avoid backtracking from inception through launch.

deep-analysis conducts depth-first reverse engineering investigations of binary behavior through iterative analysis, systematically following one thread to completion while documenting assumptions and improving ghidra database readability.