Skill UI

This sub-agent is specialized in generating precise delta specifications (delta specs). It analyzes a given proposal to identify added, modified, or removed system behaviors, creating structured requirements and scenarios to document the proposed changes accurately. It follows defined conventions for persistence across various artifact storage modes (Engram, OpenSpec, Hybrid).

This sophisticated sub-agent takes comprehensive Software Design Documents (SDD), proposals, and specifications to generate a detailed, actionable task breakdown in the format of a tasks.md file. It structures the work into distinct phases (e.g., Foundation, Core Implementation, Testing, Cleanup), providing concrete, verifiable, and small units of work suitable for immediate engineering execution and managing complex feature delivery chains.

Run optional practical diagnostics against available implementation and artifacts within Spec-Driven Development. Inspect code, run tests, build and type checks, and report findings, limitations, and next steps without mutating code.

This meta-skill audits, refactors, and standardizes the structure of existing skill definition files (SKILL.md). It ensures compliance with defined style guides, enforces best practices (Hard Rules, Decision Gates), and improves the overall quality and maintainability of LLM-based skills.

A comprehensive, black-box workflow designed for professional bug bounty hunting and penetration testing. It guides users through a 5-stage methodology (Intake, Recon, Enum, Hunt, Report), covering 19+ advanced attack playbooks (e.g., RCE, SQLi, XSS, SSRF). The skill integrates real-world exploit cases, structured payloads, and WAF bypass techniques for maximizing vulnerability discovery in web applications and APIs.

A skill for triaging issues by root cause, avoiding symptom-by-symptom fixes, and designing solutions that simplify systems. It provides hard rules for classification, test evidence, and over-engineering prevention, suitable for maintainers handling bug reports and backlogs.

A structured skill embodying Tang Jia San Shao's web fiction creation approach, covering daily updating discipline, cultivation systems with quantifiable levels, team-based battle narratives, IP lifecycle planning via six waves, and emotional anchoring from real-life experiences. Suitable for designing long-running fantasy series, managing consistent writing output, and planning cross-media IP expansion.

This skill models Tian Can Tu Dou (author of Battle Through the Heavens) web novel creative thinking, covering the 'divorce engagement' trope, progression fantasy systems, map-changing narrative loops, underdog rise golden triangle, 'terror like this' battle formulas, and IP-driven content operations.

Plan commits and PRs as reviewable work units instead of file-type chunks. Covers splitting rules, work-unit checklists, doc/test co-location, and chained PR strategies for healthy review load.

This skill provides a strategic framework for content creation within the AI Agent vertical. It emphasizes hands-on product testing, translating complex technical concepts for general audiences, community-driven distribution, and practical tutorials. Ideal for tech media creators, AI product reviewers, and content strategists aiming to build authority in the AI Agent space through verified experience and clear communication.

This skill compiles user creative intent into executable image and video prompts for AI models. It specializes in character consistency, cinematic storyboards, and reference asset mapping for R2V/I2V/T2V tasks. It ensures visual contracts, narrative flow, and production-ready specifications without replacing model capabilities.

Provides domain knowledge for compiling character/scene/prop creative settings into executable image generation prompts. Covers reference board layouts (character identity boards, sequence keyframes, environment study boards, prop study boards), composition and camera language, avoiding image moderation false positives, entity continuity, and cross-variant derivative design based on the main image. Not for video prompt compilation, pure material editing, or evaluating generated outputs.