AI & ML

impertio-studio Bundle

Use when a CesiumJS click handler hits nothing, returns the wrong object, or reports a position at the wrong height, or when building distance and area measurement tooling. Covers scene.pick, drillPick, pickPosition, globe.pick, camera.pickEllipsoid, sampleHeight, clampToHeight, ScreenSpaceEventHandler input events, selectedEntity and the InfoBox, and geodesic versus straight-line distance. Keywords: Cesium picking, scene.pick, scene.drillPick, scene.pickPosition, pickPositionSupported, globe.pick, camera.getPickRay, camera.pickEllipsoid, sampleHeight, clampToHeight, sampleHeightMostDetailed, ScreenSpaceEventHandler, setInputAction, ScreenSpaceEventType, LEFT_CLICK, MOUSE_MOVE, selectedEntity, InfoBox, Cesium3DTileFeature, EllipsoidGeodesic, surfaceDistance, Cartesian3.distance, distance measurement, area measurement, pick returns nothing, click hits nothing, picked wrong object, wrong position, position at wrong height, model floats, measurement off, how do I get the coordinate I clicked, how do I select an

jie-meng Bundle

Create new skills and prompts for any AI platform (Claude Code, Claude.ai, OpenAI, Cursor, Windsurf, etc.), modify and improve existing ones, and measure performance. Use when users want to create a skill or system prompt from scratch for any AI tool, update or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a Claude Code skill's description for better triggering accuracy.

jie-meng Bundle

Export the complete context of a multi-repo fullstack workspace, a single repo, or any user-specified directory into ONE self-contained, AI-optimized Markdown brief. Purpose: paste it into a web AI (ChatGPT, Claude.ai, Gemini) and continue architecture/design discussions without the local CLI agent, saving local tokens. Pure text: no local file or image references (paths appear as inline code text), mermaid diagrams and internet links allowed. Modes: workspace (writes into the docs repo docs directory), single-repo (writes into the repo docs directory), or explicit — when the user names a directory or subject, the user's instruction wins over auto-detection. Destination path and filename MUST be confirmed with the user before writing. Includes mandatory secret-redaction rules. Trigger: "fullstack brief", "context brief", "context handoff", "export context", "generate context doc", "brief me on this codebase", "导出上下文", "上下文简报", "生成简报", "上下文交接", "项目简报", "喂给网页AI", "网页AI上下文".

jie-meng Bundle

Vision fallback for AI models that CANNOT process images. Use ONLY after the current model fails to see a user-provided image — e.g. it says "当前模型不支持 图片输入", "无法识别图片", "无法查看这张截图", "模型不支持图片", "不支持图片", "I cannot view images", "model does not support image input", "cannot see image", or otherwise admits it cannot read the image. In that case, use this skill to analyze the image via the free GLM-4.6V-Flash API (strong OCR / text extraction) and relay the result to the user. NEVER use this skill when the current model CAN see the image — it is strictly a last-resort fallback for image recognition failures.

jie-meng Skill

Read the current screen of a connected Android device via adb. Use when the user asks to inspect or summarize what is visible on a phone/tablet screen, extract UI text or controls, debug Android app state, or mentions adb, screencap, uiautomator, current screen, or connected Android devices. When multiple devices are connected, list them with model info and let the user choose. Prefer the UI hierarchy for exact text and element metadata; fall back to a screenshot when the hierarchy is empty or the screen is visual-only. Delete all temporary screenshots and XML dumps immediately after analysis.

ruicore Bundle

Use the genmedia CLI to search, inspect, run, and manage 1200+ fal.ai model endpoints. Trigger when the user mentions "genmedia", "fal CLI", or asks to "search models", "run a model", "fetch schema", "check pricing", "upload to fal", "queue async job", "track request", or any direct interaction with the fal.ai endpoint catalog. This is the foundational skill. Every other fal.ai-related skill in this repo executes its work through genmedia commands. Use `--json` whenever the output will be parsed by an agent.

ruicore Bundle

General-purpose architecture review for any code repository. Use when Codex needs to assess concept ownership, authoritative sources, module boundaries, architecture decisions, architecture drift, duplicated concepts, scattered rules, hidden assumptions, over-engineering, architectural change surface, phased refactor planning, or how architecture exposes ownership and change boundaries to future coding agents. Do not use for ordinary code review, testing review, implementation review, framework review, style review, narrow bug diagnosis, or a standalone agent-legibility audit unless the user explicitly asks for architecture assessment.

ruicore Bundle

Use when a user wants to search, export, organize, or automate Readwise or Reader data from Codex or another terminal agent, especially when browser auth should be avoided and an access token is available.

ruicore Bundle

Review Python backend repositories at the foundation layer for target Python version compatibility, Python implementation quality, type system clarity, async/sync correctness, exception semantics, configuration and environment handling, logging and observability, packaging and dependency hygiene, Python package hygiene, testing and maintainability, and Python complexity/readability. Use when Codex needs a version-aware Python backend fundamentals review against a user-specified or inferred Python version. Do not use for architecture review, deep agent-legibility review, framework-specific FastAPI, SQLAlchemy, Pydantic, or Django review, security audit, performance deep dive, onboarding review, or business-domain review unless the user explicitly asks to include those scopes.

ruicore Bundle

Framework-agnostic repository review from the perspective of a future AI coding agent. Use when Codex needs to evaluate whether a repo is easy for agents to find, understand, modify, and safely evolve with minimal ambiguity; find discoverability problems, missing or duplicated authoritative locations, hidden conventions, cross-file invariants, ambiguous ownership, scattered change surfaces, unpredictable impact, and agent risk hotspots. This skill evaluates the agent interaction surface; use architecture-review to decide or redesign architectural ownership and boundaries. Do not use as an ordinary code-style, framework-specific, bug, performance, security, clean-code, maintainability, or generic architecture review unless the user explicitly asks for agent-legibility.

ruicore Bundle

Review how Python repositories use important ecosystem libraries and integration boundaries, automatically discovering concrete usage and evaluating it against version-appropriate practices. Use when Codex needs a Python review of framework, validation, ORM, migration, dependency injection, async, background job, HTTP client, testing, LLM/agent, observability, or runtime-library integration points. Review-only by default; do not use for broad architecture, generic Python style, full agent-legibility, security, performance, or framework-only reviews.

ruicore Bundle

Use when the user wants an AI Systems Engineering Signal Radar: researching AI systems engineering trends, AI infrastructure, agent infrastructure, production AI and LLM application engineering, agent/RAG/evals/MCP/workflow automation trends, YouTube/podcast/blog/GitHub/hiring-market signals, career skill demand for AI systems engineers, or Medium/LinkedIn/GitHub portfolio content opportunities. Optimized for a backend engineer with strong Python production experience transitioning into AI systems engineering and AI infrastructure engineering.

ruicore Bundle

Coordinate repository implementation through a master agent that discovers local rules, persists a scoped plan, delegates every production-file change to sub-agents, enforces authority boundaries, escalates material uncertainty, and closes evidence-based validation gates. Use when the user asks for delegated or multi-agent repository execution, a master/sub-agent workflow, staged autonomous implementation, parallel work with conflict control, or strict independent review of agent-written changes.

speciliam Bundle

Claude-only version of the LinkedIn early-career weekly drain — no Codex at all. Claude (Opus 4.8) is the conductor AND every worker is a fresh Claude subagent (Agent tool). Claude builds the durable run state into an isolated /tmp file, then drains last-week Entry-level software-engineer results one stage at a time by launching one fresh Claude subagent per discover/tailor/apply stage, runs a reasoning checkpoint after every stage, and reconciles the tracker at the end. Browser work uses Claude-in-Chrome / Computer Use (never the Codex Chrome plugin, never codex exec). Use when the user says `$linkedin-early-career-weekly-claude-only`, "run the weekly with only Claude workers", or wants the weekly drain done entirely by Claude with no Codex.

hugobowne Bundle

SpecStory Lore - mine your SpecStory coding histories (any agent - Claude Code, Codex, Cursor, Gemini, and more) into a persistent corpus, surface your reproducible workflows with corroborated evidence, and interactively forge the chosen ones into skills installed across all your agent harnesses. Use when the user wants to turn past AI coding sessions into reusable skills, asks "what could I make into a skill", "mine my lore", "forge skills from my history", or points at a .specstory/history directory.

hugobowne Bundle

Build, extend, AND query a persistent LLM-maintained wiki for any research topic. Conversational entry point that routes between four modes — query (read-only Q&A against existing research with optional Q&A save-back), init (new topic ingest), append-deep (add sources to existing topic via short rounds), and append-trusted (add a single user-vouched file to existing topic). Ingests from the Second Brain (Obsidian + Readwise + NotebookLM + GitHub repos + web seeds + user-dropped PDFs) and maintains a wiki layer (per-source pages, entities, concepts, comparisons, overview, synthesis, open questions, contradictions). Use for any research interaction — first-time research on a topic, "what do I have on X", "load my research on Y", "add this PDF to my research", "deep dive on Z", "pull together my notes on Y", "extend my research with this file". Trigger on the phrases above plus "search my research", "use my research", "find sources about X".

hugobowne Bundle

Draft a goal+rider document pair to brief the next agentic turn on a project. Two documents — a ≤4000-char goal (the spine) and an unbounded rider (the prescriptive detail with eleven phases and depth-tests-first discipline). Run when user says "draft a goal", "new goal", "write a goal+rider", "goal+rider for the next agentic turn", "rider for X", or anything in the shape of "brief the next agent on Y".

hugobowne Bundle

Refine a prompt by interviewing the user's intent from multiple angles, generating multiple variations at different magnitudes of change, and scoring each against an editorial rubric defined before the refinement starts.

hugobowne Bundle

Analyzes threads for friction points where the model struggled or user corrected it. Produces actionable improvements for AGENTS.md, skills, tools, and codebase. Use when asked to review a thread, do a postmortem, or analyze what went wrong.

hugobowne Bundle

Research a topic and build a beautiful, magazine-style PDF explainer written for a specific kid — designed cover, sections, cards, comparison tables, callouts, real photos, and a closing prompt. Use this whenever Tim wants to make a guide, explainer, handbook, or "writeup" for one of his kids (e.g. "make James a guide to volcanoes", "build a magazine explainer about how rockets work for my daughter", "turn this chat into a nice PDF for my son", "research X and make it into something my kid can read"). Triggers on "explainer", "magazine-style", "field guide for my kid", "make a PDF for [child]", or any kid-facing illustrated explainer — even if Tim doesn't say "magazine". Produces a print-ready PDF via headless Chrome and always audits the prose against PROSE.md and the stop-slop skill before finalizing.

hugobowne Bundle

Apply rich formatting to Notion pages - toggles, callouts, headers, and styled text. Use when creating or formatting Notion documents via the Notion MCP server.

hugobowne Bundle

Use when you have an idea for a data story and want a publication-quality chart back. The agent finds an authoritative public dataset, drafts three chart-type variants in parallel, picks the strongest, then iterates against a design checklist and a chart-design evaluator until the chart actually carries the story.

hugobowne Bundle

Use when a user wants an agent to summarise every video in their YouTube Watch Later playlist, publish one secret GitHub gist per summary with gh, and return channel/title/gists.sh summary links. Browser-tool agnostic; assumes the agent can operate the user's logged-in browser/session.

tss-blue-ai-lab Bundle

Audit rules and skills for semantic contradictions, redundancy, and staleness, then interactively resolve with the user. Use periodically when agent performance degrades or after adding many rules/skills.

tss-blue-ai-lab Bundle

Optimize an agent skill through bounded repeated trials, binary evaluations, targeted prompt mutations, and auditable keep/discard decisions. Use when asked to optimize, benchmark, evaluate, or systematically improve an existing skill. Produces an improved candidate skill, run artifacts, a results log, and a mutation changelog without overwriting the original.

tss-blue-ai-lab Bundle

Review uncommitted git changes for bugs/regressions via Codex MCP and present a structured report. Use when asked to review local changes or find bugs in current work. Requires Codex MCP to be configured.

tss-blue-ai-lab Bundle

Capture a full DevTools-protocol trace of any browser automation — CDP firehose, screenshots, and DOM dumps — then bisect the stream into per-page searchable buckets. Use when the user wants to debug a failed run, audit network/console/DOM activity, attach a trace to an in-progress session, or feed structured per-page summaries back into an agent loop so its next iteration learns from the last one.

tss-blue-ai-lab Bundle

Create well-structured agent skills following cross-platform best practices. Use when building a new skill, improving an existing one, or creating automation for a recurring workflow.

tss-blue-ai-lab Skill

Craft and optimize prompts for better AI output. Improve clarity, structure, and effectiveness of prompts for Claude Code, Codex, and other AI tools. Use when: getting poor results from AI, building reusable prompt templates, or creating skills and agents.

tss-blue-ai-lab Skill

Design database schemas, write migrations, and model relationships. Use when: starting a new project that needs a database, adding tables, designing relationships, or optimizing queries.

tss-blue-ai-lab Skill

Coordinate specialized agent roles across complex tasks using dependency-aware planning, explicit ownership, structured handoffs, safe integration, and evidence-based quality gates. Use when work has multiple independent streams or genuinely benefits from isolated expertise; fall back transparently when the runtime cannot provide separate workers.

cryptopafi Bundle

Run a 3-model advisor council for strategic decisions, architecture choices, and plans. Use when Pafi asks for /council, council review, or multi-model decision audit. ANTI-PATTERN: Do not use for code artifact quality review; use audit-pro.

cryptopafi Skill

Run a comprehensive Codex session handoff: save important findings to Cortex, update impacted project cards and sync them to Cortex, write durable findings to LLM-Wiki when reusable, append Genie updates, write a session summary, preserve pending tasks, check agents/modes, sync relevant git repos when available, and return a final report. Use when Pafi says handoff, /handoff, save session, wrap up everything, session end, or prepare next session. ANTI-PATTERN: Do not use for quick summaries, single-file saves, or WISH Step H task delegation.

cryptopafi Skill

Consensus architectural planning — a GPT-5.6 Sol/high planner drafts RALPLAN-DR, two parallel GPT-5.6 Sol/max auditors critique it, and a Sol/max reconciler produces the final verdict. Output saved to ~/.nexus/workspace/plans/{name}.md. Default SHORT mode; pass '--deliberate' for pre-mortem + expanded test plan. Auto-invoked by TECH when Source Plan is missing. Triggers: 'ralplan', 'consensus plan', 'plan this properly', 'architect this', 'plan with review', 'planifică asta', 'plan consensual'. NOT for: prompt optimization (promptforge-scope), research (delphi), Q&A.

cryptopafi Skill

Run one Self-Optimization Loop queue item through audit, validation, auto-mode or approval, patch application, Cortex save, and manifest update. Use when Pafi asks for sol-cycle or one SOL optimization cycle. ANTI-PATTERN: Do not use for general prompt optimization; use opt.

cryptopafi Skill

Create, convert, assemble, or build portable NexusOS procedures, skills, and plugin bundles using the canonical FORGEBUILD v2.1 procedure. Use when Pafi says forgebuild, forge/build a procedure, create a reusable workflow, turn a procedure into a skill, build a plugin/skill bundle, or standardize a skill across Codex/Claude runtimes. ANTI-PATTERN: Do not use for one-off task execution, ordinary code edits, or prompt optimization.