AI & ML

formars0309-cloud Bundle

Discipline for producing clinical content a clinician will act on - drug doses, target values, thresholds, management steps, guideline recommendations, anesthesia/ICU/perioperative plans, differential and workup summaries. Load BEFORE drafting, not after. Use when the user asks about 마취, 용량, 목표치, 프로토콜, 가이드라인, 약제, 환자 관리, 술기, or says "정리해줘"/"알려줘" about a clinical topic; and for "what dose", "what target", "how do I manage", "summarize the management of", "protocol for". Enforces guideline lookup before drafting, a source label on every number, format rules that stop conditions being flattened, and cross-model review before delivery. Companion to [[citation-verify]], which checks references after text exists; this one governs the writing itself.

okf-memory Bundle

Maintain persistent, domain-neutral project memory for AI agents using Open Knowledge Format (OKF) v0.2 bundles and the deterministic okf Go toolchain. Use whenever project knowledge, decisions, runbooks, research, client notes, or domain discoveries must survive conversational resets.

oswaldhill Bundle

Audit and safely align installed IDEs and agents against a shared Skills repository and a declared MCP endpoint profile. Verify installation evidence, Skills links, MCP configuration, and live endpoint capabilities as four separate states, and repair or roll them back safely. Use when checking or fixing IDE/Agent Skills links, MCP configuration, endpoint liveness, capability coverage, legacy MCP channel migration, or connection scale.

alharkan7 Skill

Agent software engineering behavior and architectural principles.

alharkan7 Skill

Use this skill when the user wants to turn a scientific research paper, dataset, or text document into a stunning, interactive "Scrollytelling" web application.

synthesisengineering Bundle

Combine multiple LLM responses into a single unified document. Use when asked to combine responses, merge outputs, synthesize responses, unify documents, or consolidate multiple AI-generated answers into one comprehensive result.

synthesisengineering Bundle

Plan, record, and audit text provenance across hosted, local, and open-weight model workflows. Use for text provenance, watermark capability checks, local-model generation, reproducible AI-assistance records, mixed-authorship lineage, text-integrity audits, authorized detector results, and claims about what a provenance signal can or cannot prove. Do not use to defeat provider marks, evade detectors, or disguise AI authorship.

synthesisengineering Bundle

Audit, install, and verify a synthesis ecosystem across multiple AI agent runtimes. Use for Claude Code and OpenAI Codex parity audits, AGENTS.md and CLAUDE.md instruction migrations, skill or plugin deployment checks, lifecycle-hook health, Mac bootstrap validation, active-project handoffs, post-compaction recovery, and any request to make synthesis project management portable between agent clients.

synthesisengineering Bundle

Two-category disclosure governance for people who publish under their own name while handling confidential work. Distinguishes published-precedent facts (deliberately public biography an agent may restate) from unapproved disclosures (anything learned from private context), with a precedent ledger, surface classes, five decision tests, and git-hook enforcement. Use when asked about: disclosure policy, confidentiality guardrails, can I name this company, precedent ledger, public bio names, publication surface, unapproved disclosure, name allowlist.

synthesisengineering Bundle

Configure an LLM as a senior technical advisor for software development and engineering. Use for technical advisor, tech setup, configure advisor, technical assistant, architecture review, code review guidance, and engineering decisions.

synthesisengineering Bundle

Run bounded, differently-shaped-agent adversarial review against the principal's outcome, with artifact-complete rounds, production-topology handoffs, explicit concessions, a fail-closed finding ledger, sufficiency rulings, and independent post-publication acceptance. Use for adversarial review, cross-agent review, red-team collaboration, review rounds, finding-ledger work, or reviewer handoffs.

synthesisengineering Bundle

Lightweight project management system designed for human-agent collaboration, optimized for context preservation and cross-agent coordination across sessions. Use when asked to: project management, project setup, project tracking, synthesis project, manage project, set up project, project structure, session protocol, parallel root sessions, advisory locks, cross-agent coordination.

beatapi Bundle

Use when a user asks an agent to call BeatAPI Model, Social Data, or Workflow capabilities. Prefer bundled MCP tools when available or the official CLI as a fallback; covers text, image, video, social-data actions, Effects, Music Video, Ecommerce Video, Video Analysis, Realtime sessions, task monitoring, usage, webhooks, and API errors.

synthesisengineering Bundle

Validate, convert, and author content for Google's Open Knowledge Format (OKF v0.1) — the markdown-plus-YAML-frontmatter knowledge-bundle spec announced 2026-06-12. Includes a conformance validator (the checker Google's own OKF repo ships none of), a converter that backfills OKF frontmatter onto an existing markdown corpus idempotently, a config-driven seven-point metadata-consistency checker, and the proven repo-by-repo conversion procedure. Use when adopting OKF for an "LLM wiki" style knowledge base, auditing conformance, checking frontmatter/body drift or taxonomy consistency, or converting a corpus of markdown notes/docs into a conformant, agent-readable bundle.

synthesisengineering Bundle

Guide for setting up and configuring LLM projects across platforms. Use when setting up a Claude Project, creating a ChatGPT GPT, configuring a Gemini Gem, setting up an AI assistant with custom instructions and knowledge, or configuring any LLM platform with AI Knowledge content.

synthesisengineering Bundle

Profile a computer, compare Ollama, LM Studio, llama.cpp, and MLX-LM, recommend local open-weight model artifacts that fit its real memory and storage, install approved artifacts through deterministic managed-runtime adapters, update installed Ollama models with before-and-after identity receipts, maintain a privacy-safe per-machine inventory, and verify local inference. Use for: local models, open weights, Ollama, LM Studio, llama.cpp, MLX model selection, model updates, which model fits this Mac or PC, install Qwen/GLM/Kimi/DeepSeek locally, hardware profile for LLMs, model inventory, local inference benchmark.

synthesisengineering Bundle

Compose and send honest agent correspondence across Slack, email, and other channels. Defines principal-direct, assistant, and bot lanes; the voice axis (chief-of-staff personas speak as the principal, executive-assistant personas speak as themselves); review-depth governance; persona configuration; disclosure signatures; and compose/send gates. Use for agent correspondence, sending on a principal's behalf, message signatures, disclosure lanes, persona registries, agent branding, agent voice, third-person agent messages, standing-direction sends, ghostwriting disclosure, or outbound-message gates.

synthesisengineering Bundle

Evidence and provenance discipline for AI-agent output — the truth-side companion to synthesis-anti-shortcuts' effort-side discipline. A catalog of grounding rules: never record imagined events, quote only what a tool surfaced, treat context files and memories as caches to re-verify before propagating, name the layer a config claim describes, read the evidence in hand before theorizing, count the corpus before generalizing, prove absence with a positive control and bounded reads, never complete truncated output, and validate paths before writes and deletions. Use when asked to: grounding discipline, verify claims, evidence check, provenance check, anti-confabulation, absence claim, negative finding, zero results, cache vs truth, count the files, truncated output, verify paths, safe deletion, is this grounded.

nahid-sparktales Bundle

Build an MCP server, or bring an existing one into a project — choosing the transport, deciding which tools, resources and prompts to expose, keeping reads separate from writes, handling auth and credentials, and defining what failure looks like to the model. Use when writing an MCP server, wrapping an internal system as one, or evaluating a third-party server before wiring it in. Not for designing the individual tool signatures inside it (tool-design), and not authorization to install, configure or run any server against real systems.

nahid-sparktales Bundle

Build an eval suite that can actually detect a regression — cases pulled from real traffic, graders that check properties rather than vibes, a recorded baseline, and per-case diffs in both directions. Use before claiming a prompt, model or agent change is an improvement, when agent behaviour must not regress, or when someone reports "it seems better" after eyeballing a handful of outputs. Not for tracing what one run did (llm-observability), not for testing deterministic code, and never as evidence that behaviour the suite does not measure is safe.

nahid-sparktales Bundle

Design the tools a model calls — names, parameter shapes, what a result returns, and error text written as an instruction the model can act on. Use when adding or reshaping a tool or function an LLM invokes, when an agent keeps calling the wrong tool or passing malformed arguments, or when reviewing a tool surface someone else defined. Not for building the MCP server that hosts the tools (mcp-design), not for forcing a model's final answer into a schema (structured-output), and not for prompt wording outside the tool definition.

nahid-sparktales Bundle

Scope an agent or subagent before it is built — the one job it owns, the smallest tool set that closes that job, what it must never do, and the evidence it has to return. Use when adding an agent, subagent or automated role to a system, when deciding which tools it gets, when an existing agent loops, over-reaches or reports work it did not do, or when reviewing someone else's agent design. Not for wording the prompt itself, not for deciding what occupies its context window, and it never grants an agent permission it did not already have.

nahid-sparktales Bundle

Handle requested Agent Dispatcher work with direct execution or prepared specialist guidance. Includes named roles, context inspection and controls.

fruitflyworld Bundle

Win a free Fruit Fly World Genesis Passport by planning a foraging route. Time is cut into one-hour windows; every agent in a window solves the SAME map, and the single free slot goes to the BEST SCORE when the clock hits zero. Install this skill into an AI agent and it can pull the brief, search routes offline, sign with its own wallet, and enter every hour without a human. Use this whenever the user asks to "win the Foraging Hour", "mint a Fruit Fly World Passport free", "run the fly arena", or "play fruitfly.world".

lnyo-cly Bundle

Use this skill when helping users build applications with AI4J in their own Java or Spring Boot projects, including first chat, streaming, tool/function calls, MCP, RAG, memory, Agent runtime, Coding Agent CLI embedding, FlowGram integration, provider configuration, dependency selection, and troubleshooting. It guides beginner-friendly app scaffolding, secure environment-variable configuration, smallest useful AI4J module selection, runnable examples, and verification steps. For AI4J repository maintenance, follow the repository AGENTS.md and Harness Anything task workflow instead of this user-facing app builder skill.

lnyo-cly Skill

Ask the application user concise structured questions when missing information blocks a safe AI4J agent action.

hraness Bundle

Set up, evolve, or operate a hraness/wordcell local-first Markdown knowledge base for coding-agent memory. Use when a user asks to design Wordcell conventions or a recurring Wordcell ritual; search or query a Wordcell or Obsidian vault; load or recover repository context, plans, decisions, concepts, backlinks, semantic search, or Git provenance from an earlier coding session; save, clip, scrape, or archive a URL, article, social thread, signed-in browser page, or PDF as auditable Markdown; create or update a durable plan in the vault; or refresh, check, percolate, and maintain its knowledge graph. Do not use for generic web research, generic PDF reading, or ordinary planning that will not use a hraness/wordcell vault.

synthesisengineering Bundle

Slack channel sync protocol for AI-assisted workflows. Reads channels and threads via Slack MCP, saves to local transcript files in workspace-scoped repos, and updates person-scoped daily action plans. Handles mid-day re-syncs with thread staleness detection. Use when asked to: slack sync, sync from slack, check slack, read channels, sync messages, sync transcripts, what's new on slack.

synthesisengineering Bundle

Cross-provider model-tier convention for agentic work: three role labels (judgment, routine, bulk — formerly frontier, efficient, light) resolved to current model IDs per provider in tiers.yaml, so skills, project docs, and memory never hardcode model names. Also carries the role-selection rule: route by whether the CAUSE is known, not by how small the task looks — a symptom report is diagnosis and belongs in judgment even when the subject is one file. Use when asked about: model tiers, which model, model selection, judgment model, routine model, bulk model, frontier model, efficient model, switch models, model equivalents across providers, update model table, which effort level, low effort, wrong model for the task.

synthesisengineering Bundle

Prepare a principal for any meeting the way a wise chief of staff would: weigh 60+ factors across the meeting, participants, principal's position, knowledge, and risk; model the readers before drafting; deliver a dense, scannable pack with a capture half; then debrief the transcript into decisions, commitments, and reader-profile updates. Use for 1:1s, reviews, forums, external meetings, interviews, and post-meeting follow-through.

synthesisengineering Bundle

Route a request to the correct synthesis engineering, coding, writing, project-management, knowledge, operations, or agent-governance skill while keeping specialist metadata out of Codex's bounded prompt. Use when a task appears to match a synthesis workflow but the user did not name the exact skill.

nahid-sparktales Bundle

Decide what an agent should remember, which layer holds it, who it is scoped to, and how a stale or contradicted memory is detected and retired. Use when an agent forgets something across sessions, when a memory or persistent-context feature is being designed, or when stored memories have grown noisy, wrong, or are leaking between users. Not for retrieval over a document corpus, not for prompt or context-window packing, and not for conversation transcript storage.

nahid-sparktales Bundle

Pick a model per job and degrade sensibly when one fails — a quality bar per call site, candidates compared on the same task set, a readable routing rule, and an explicit retry-versus-fallback path with pinned model ids. Use when cost or latency has become a problem, when adding a cheaper or larger model to an existing system, or when a fallback fires silently and quality drops without anyone noticing. Not for prompt authoring, not for retrieval tuning, and not for capacity or infrastructure planning.

nahid-sparktales Bundle

Build and fix retrieval that actually returns the right passage — structure-aware chunking, one pinned embedding model, lexical plus vector search fused, reranking, and a recall measurement that is run separately from the generator. Use when a RAG system answers wrong or vaguely, when an index is being designed or reindexed, or when someone proposes a prompt change to fix what is really a retrieval miss. Not for prompt or output-quality work once the right passage is already in context, not for agent memory design, and not for choosing a vector database.

synthesisengineering Bundle

Systematic fact-checking for articles, blog posts, news content, and AI-synthesized material. v2.0 adds nine new protocol sections covering nested attribution, paraphrase drift, composite quotes, position-shifting, source-translation drift, URL rot vs hallucination, AI-generated synthetic sources, citation laundering chains, and tool-specific hallucination patterns by LLM family. Use when asked to: fact-check, verify claims, verify sources, check accuracy, citation verification, review factual accuracy, validate references, audit AI-summarized content.

synthesisengineering Bundle

Manifest-driven email inbox cleanup across three tool stacks: iCloud / generic IMAP via Python + YAML rules; Microsoft 365 + outlook.com via Mail.app AppleScript; Gmail via workspace-mcp Gmail API and server-side filters. Engine is public; per-user rules live privately at ~/.synthesis/inbox-cleanup/. Ships with prompt-injection defenses (sanitization module + adversarial test fixtures) for any LLM-augmented path. Use when asked to: clean up inbox, sweep email, categorize senders, build email rules, archive promotions, set up Gmail filters, build email cleanup automation.