Agent Skills Hub

reflect-appworld-failure

Analyze AppWorld task failures to extract specific API patterns and generate actionable playbook bullets with concrete code examples.

root-cause-tracing

Systematic debugging skill to trace errors backward through call stacks, identify original triggers, and implement layered defenses instead of patching symptoms.

browser-debugger

Comprehensive UI testing, visual fidelity analysis, and browser debugging using Chrome DevTools MCP and AI-driven vision models.

ring:dev-sre

Gate 2 development cycle skill that validates observability implementation, including structured logging, OpenTelemetry tracing, and instrumentation coverage, without modifying code.

novelty-check

Verify research idea novelty against recent literature. Use when user says '查新', 'novelty check', or needs to confirm if a method is original.

doncheli-visual-test

Perform automated visual regression testing by comparing UI screenshots against established baselines to identify layout shifts, color changes, and rendering regressions.

design-review

Designer's eye QA: detects and automates fixes for visual inconsistencies, spacing, hierarchy, and UI polish issues. Iteratively verifies with before/after screenshots.

deepscan-monitor

Manage long-running PapersFlow DeepScan research workflows with asynchronous monitoring, live progress tracking, and automated report generation.

skill-vetter

Security-first vetting protocol for AI agent skills. Detects red flags like credential theft, obfuscated code, and unauthorized data exfiltration before installation.

toh-framework

Security-first auditing framework for AI-generated code. Provides multi-level protection including hardcoded secret detection, dangerous pattern identification, and comprehensive vulnerability audits for modern web applications.

sound-effects

Generate professional-grade sound effects from text descriptions. Create audio textures, cinematic impacts, UI sounds, and ambient environments with precision control over duration, looping, and prompt adherence.

youtube_moderation_prototype

A prototype skill for automating YouTube live chat moderation using pattern-based detection for spam, toxic content, and rate limiting, optimized for testing agent reliability before deployment.