Loading...
Loading...
Found 796 Skills
Use when you need to generate/update test plans, test cases, test suites or test reports during the verification phase of Spec Pack, and require strict access control, traceable artifact storage, and no unauthorized routing.
Use when developing or modifying visual elements (components, pages, layouts, styling). Covers Playwright screenshots, Compodium component preview, responsive/dark-mode verification, and the iteration loop.
Browser automation CLI for AI agents. Use for website interaction, form automation, screenshots, scraping, and web app verification. Prefer snapshot refs (@e1, @e2) for deterministic actions.
Instant visual verification via screenshots. For quick checks like 'does button look blue', 'is layout centered', 'header look right on mobile'. Fast alternative to formal testing - just look and confirm. Use when user wants visual inspection without creating test files.
Implement approved technical plans phase-by-phase, including code changes, verification, and progress updates directly in the plan file. Use this whenever the user asks to execute a plan, implement from a plan document, continue plan-based work, resume from checked/unchecked tasks, or "do phase 1/2/3" from an implementation plan in `rpi/plans/`, even if they do not explicitly name this skill.
Validate email addresses before campaign sending. Takes a contact CSV, validates each email via a verification provider, removes invalid/do_not_mail/abuse/unknown addresses, and optionally cleans them from sequencer campaigns. Outputs a verified CSV ready for campaign-sending. Fits between email-generation and campaign-sending in the pipeline. Triggers on: "verify emails", "validate emails", "email verification", "clean emails", "check emails before sending", "remove bad emails", "email hygiene".
A specialized skill for generating high-quality illustrations for academic papers, supporting two output formats: (1) LaTeX/TikZ code: Suitable for structured diagrams such as system architecture diagrams, data flow diagrams, and geometric schematic diagrams, which can be directly embedded into papers; (2) draw.io XML: Suitable for highly decorative diagrams such as technical roadmaps, scientific research display diagrams, and academic presentation illustrations, supporting gradient colors, shadows, and free layout, which can be opened and edited at app.diagrams.net. Supports the above two output formats with a unified workflow: Analyze input (copy/image/paper) → Drawing instructions → Code generation → Compilation verification → Full-score delivery. It automatically identifies the field of the paper and designs illustrations as an expert in that field. Use when the user asks to: 画论文图、画架构图、画流程图、画示意图、 LaTeX画图、TikZ画图、论文配图、生成画图指令、复刻图片、 画图代码、学术论文图、画系统架构、画协议流程、论文插图、tikz diagram、 latex figure、根据论文画图、画个图、帮我画图、生成tikz、论文tikz、 根据文案画图、照着图片画、复刻这张图、技术路线图、科研架构图、 学术汇报图、drawio、draw.io、路线图、研究框架图、技术方案图。
Use before claiming any task is complete, any feature works, or any bug is fixed - enforces fresh verification evidence through a 5-step HARD-GATE protocol that prevents false completion claims
Claiming work is complete without verification is dishonesty, not efficiency. Use when ANY variation of success/completion claims, ANY expression of satisfaction, or ANY positive statement about work state.
Debugging and Root Cause Localization for AscendC Operator Precision Issues. Used when operator precision tests fail (such as allclose failure, result deviation, all-zero/NaN output, etc.). Process: Error Distribution Analysis → Code Error-Prone Point Review → Experimental Isolation → printf/DumpTensor Instrumentation → Fix Verification. Keywords: precision debugging, precision issue, result inconsistency, error localization, allclose failure, output deviation, NaN, all-zero, precision debug.
Decode and inspect JSON Web Tokens (JWTs) without verification. Use when the user provides a JWT string and wants to see its header, payload, or claims — e.g. "decode this JWT", "what's in this token", "inspect this JWT", "show me the claims", "parse this token". Also triggers on raw JWT strings (three base64url segments separated by dots).
Launch multiple sub-agents in parallel to execute tasks across files or targets with intelligent model selection, quality-focused prompting, and meta-judge → LLM-as-a-judge verification