AI & ML

outlinedriven-odin-claude-plugin Bundle

Use when the user wants to batch-close resolved or outdated tracker items. The agent never closes tracker items itself. Not for individual closure or items still under active work.

outlinedriven-odin-claude-plugin Bundle

Use when asked to audit an agent or AI feature for agentic-experience quality (AX review, agent-native critique, trust question). Not for source or remote-system changes.

outlinedriven-odin-claude-plugin Bundle

Use when settled conversation decisions need synthesis into an agent-ready implementation spec, stopping before publication. Not for turning plans into tickets: use to-tickets.

nordeim-openclaw-curated-skills Bundle

Transcribe audio files to text using OpenAI Whisper. Supports speech-to-text with auto language detection, multiple output formats (txt, srt, vtt, json), batch processing, and model selection (tiny to large). Use when transcribing audio recordings, podcasts, voice messages, lectures, meetings, or any audio/video file to text. Handles mp3, wav, m4a, ogg, flac, webm, opus, aac formats.

autohandai-community-skills Bundle

Error handling guidelines for NeMo-RL. Covers exception specificity, minimal try bodies, and else blocks.

autohandai-community-skills Bundle

Describes the agent skills shipped with NemoClaw and how to access them by cloning the repository. Use when users ask about AI agent support, coding assistant integration, or the .agents/skills/ directory. Trigger keywords - nemoclaw agent skills, ai coding assistant, cursor, claude code, copilot.

autohandai-community-skills Bundle

Implementing Diamond Model Analysis

autohandai-community-skills Bundle

Run, monitor, analyze, and debug LLM evaluations via nemo-evaluator-launcher. Covers running evaluations, checking status and live progress, debugging failed runs, exporting artifacts and logs, and analyzing results. ALWAYS triggers on mentions of running evaluations, checking progress, debugging failed evals, analyzing or analysing runs or results, run directories or artifact paths on clusters, Slurm job issues, invocation IDs, or inspecting logs (client logs, server logs, SSH to cluster, tail logs, grep logs). Do NOT use for creating or modifying evaluation configs.

autohandai-community-skills Bundle

Query and browse evaluation results stored in MLflow. Use when the user wants to look up runs by invocation ID, compare metrics across models, fetch artifacts (configs, logs, results), or set up the MLflow MCP server. ALWAYS triggers on mentions of MLflow, experiment results, run comparison, invocation IDs in the context of results, or MLflow MCP setup.

tmolavi Bundle

Use these skills when you need to fine-tune the database engine settings, manage extensions, or optimize the columnar engine for better analytical performance.

tmolavi Bundle

Use these skills when you need to discover and manage PostgreSQL extensions or fine-tune engine-level settings such as memory allocation and server configuration parameters.

jonathaneciana17-maker Bundle

Create, edit, validate, and run Agent QA tests, suites, and hooks through MCP or CLI while preserving canonical IDs and schema contracts.

axidify Skill

Use Playwright MCP to inspect a running web app when the user reports a visual or interaction issue but cannot describe it precisely. Use when they mention layout, spacing, wrong copy, button behavior, hydration flash, or "something looks off" on any page or route.

claude-ai-tools-ventura-county Skill

Restructure multi-step output into a single linear, numbered, step-by-step plan with branches as indented sub-bullets — concise but detailed. Use this whenever a response (yours or another agent's) contains multiple steps, instructions, or actions scattered across verbose prose, split across sections, or buried in explanations. Trigger when the user says "linear", "linearize", "make this step-by-step", "give me a plan", "what do I actually do", "put this in order", or pastes a verbose response and asks for the actionable sequence — including completion/status messages whose remaining work is scattered across "what I didn't do", "open items", and "next steps" sections. Also self-trigger when about to give 3+ procedural actions the user must execute themselves — but not for option comparisons, conceptual explanations, or summaries with no remaining work to execute.

claude-ai-tools-ventura-county Bundle

Starts and maintains a manual session attention firewall for Claude Code. Invoke only with /btw when the user explicitly wants focus mode, no side quests, or a maximum of three active tasks. Create one append-only Markdown log per BTW focus session, capture the opening user prompt plus repository and branch context, and park assistant-initiated off-task findings as Action, Review Later, or Interesting instead of surfacing them in assistant prose. After activation, treat plain-language "btw check" as an in-session re-anchor rather than a new invocation.

claude-ai-tools-ventura-county Skill

Distill long, verbose, or repetitive output—especially from Claude or other AI agents comparing options, tradeoffs, and recommendations—into a short, deduplicated bottom-line summary. Surface the real decision, a clear recommendation, and any choice that would be hard to reverse later. Trigger when content is decision- or tradeoff-oriented and the user asks for the bottom line, TLDR, gist, takeaway, to cut to the chase, summarize cleanly, says it is too long/verbose/repetitive, or asks what they actually need to decide or might regret. Also self-trigger when prior agent output is long-winded, hedged, or repetitive and a decision needs to be pulled out clearly, even without those exact words. Do not trigger when the user wants procedural steps, commands, or an execution sequence; that is /linear's job, not this skill's.

claude-ai-tools-ventura-county Bundle

The EXECUTOR of the optimization pair: run a bounded, self-verifying loop that measurably improves a single file, only AFTER a measurable contract already exists. Fire when a scalar metric, a strong correctness oracle, and a fixed budget are ALREADY defined — typically handed off by baseline-spec, or when the user says they have a benchmark/eval and a baseline — or when the user explicitly invokes the loop: 'run the auto-improve loop,' 'let it iterate until it's faster,' 'hill-climb this,' 'auto-tune against my benchmark,' 'Karpathy-style optimization loop.' For a cold-start request that has NOT yet named a number, gate, and stop — 'optimize this,' 'make it faster,' 'make it cleaner,' 'tune this prompt' — do NOT run this loop; defer to baseline-spec to define the contract first. Re-gate on the three preconditions before every run; if any is missing, refuse and name it. Skip entirely when the goal is subjective with no number behind it.

claude-ai-tools-ventura-county Skill

Pressure-test an option set before the operator picks. Catches an AI agent — this one included — conflating two separate problems into one menu, or quietly narrowing the field to a false dilemma that skips the boring option. Borrows debug-mantra's falsify-the-hypothesis rigor (disprove each option before trusting it) then ponytail's minimalism bias (of what survives, favor the smallest) to surface the option nobody named because it wasn't impressive enough. Trigger when an agent — this one included — or a user presents two or more options for a non-trivial engineering decision, especially when every option involves new abstractions, infrastructure, or a rewrite; when the framing sounds like "you have to either X or Y," "there's no way around it," or "we'd need to..."; or when the user asks "is that really the only way," "feels like you're overcomplicating this," "are these actually the same problem," or "what's the simplest version that still works." Also self-trigger internally before presenting any option m

claude-ai-tools-ventura-county Skill

The DEFINER and entry point of the optimization pair: turn a vague "make it better" into a measurable contract before any optimization or eval begins — a metric, an un-gameable correctness oracle, a budget, and a captured baseline. Fire FIRST, before auto-improve, on any cold-start optimization request that has named no number, no correctness gate, and no stopping rule: "optimize this," "make it faster / cleaner / tighter / more engaging," "tune this prompt," "shrink this," "improve the copy." This skill runs before the executor; do not let the loop fire on an undefined request. Default to a one-shot spec sheet that names the metric, the oracle, the budget, and the single missing fact; drop into one-question-at-a-time interrogation only if the operator engages. Refuse to optimize until all three pillars are defined. Skip only when the metric, oracle, and budget are already specified — then hand straight to auto-improve.

claude-ai-tools-ventura-county Skill

Run a brief but meaningful decision reset before committing to a plan, recommendation, or next step. Use this when the user is about to make a choice, accept an LLM recommendation, commit to a direction, or push forward on a plan that may deserve a quick 360 re-evaluation first. Especially useful when assumptions may be hidden, the problem may be framed too narrowly, urgency may be driving the choice, or the path may be hard to reverse later.

echo-aloha Bundle

Build and audit PFC2D/PFC3D 5.0 model foundations for asphalt cases, including lifecycle, domain, balls, walls, clumps, groups, ranges, top-level cmat, staged saves, and source-version purity.

echo-aloha Bundle

Select, probe, calibrate, and audit the Burger viscoelastic contact model for PFC 5.0 asphalt-mixture cases, including Maxwell-Kelvin parameter roles, units, temperature/rate dependence, PFC5 cmat assignment, timestep checks, creep/recovery validation, and PFC5-only syntax boundary checks.

kalabint Skill

Read OR write the project's session-handoff doc. WRITE mode updates it so a fresh agent can pick up the work (what was built, in-flight files, open questions). LOAD mode reads the doc AND the files it references so YOU get up to speed at the start of a session. Use when the user says "update handoff", "write a handoff", "hand off", or signals a session is wrapping up (→ write); or "read the handoff", "load the handoff", "get up to speed", "catch me up", or pastes the doc path at the start of a session (→ load).

kalabint Bundle

Audit the project's persistent agent memories for backfire-risk, stale/superseded state, contradictions, redundancy, and bloat. Lists problems for the user to disposition - does NOT auto-correct. Surfaced on a 14-day cadence by a SessionStart hook. Use when the user asks to audit/review/clean up memories, or when the overdue reminder fires.

kalabint Skill

When asked to estimate something unstated, produce a legible correctable evidence chain rather than refusing or fabricating a number.

kiyoraka Bundle

Auto-triggers when user asks for a Midjourney or NijiJourney image prompt, when creating visual art prompts, or when user says 'midjourney prompt', 'niji prompt', 'create prompt', 'create a prompt', 'image prompt', 'generate prompt', 'draw this', 'reference sheet', 'make an image of'. Generates optimized AI image prompts with composition-aware framing.

kiyoraka Bundle

MUST use when the user wants to RENDER or GENERATE an actual image from a description (not just craft prompt text). Triggers on "render an image", "render this", "generate image", "render image", "make an image of", "render the prompt", "render this prompt". Builds the prompt from what the user types (subject-agnostic, no built-in persona), optionally pulls a saved reference from the Library System, confirms cost, then calls the OpenAI gpt-image API and saves a PNG.

kiyoraka Bundle

MUST use when the user wants to RENDER or GENERATE an actual video / MP4 from a description, or ANIMATE a local image. Triggers on "render a video", "generate video", "make a video", "create a video", "animate this", "animate this image", "bring this to life". Builds the motion prompt from what the user types (subject-agnostic, no built-in persona), optionally pulls a saved reference from the Library System, supports text-to-video and image-to-video, confirms cost (video is expensive), then calls the Seedance API (async submit/poll/download) and saves an MP4.

igoingdown Bundle

用 Logfire MCP 做生产可观测性的看板运维与根因分析。四类能力：(1) 在现有 Logfire dashboard 里增/改/删 panel（Perses JSON 结构）；(2) 用 /loop 定期巡检看板与告警、对比基线、分析异常根因；(3) 按 trace_id 下钻捞全链路日志、定位根因并给出修复方案；(4) 基于历史数据做频率/影响面/SLI/趋势的量化分析，判断问题是否值得修。触发词：Logfire 看板/面板/panel/dashboard、加监控、定期巡检、监控告警、读告警群、根因/排查、trace 下钻、捞日志、值不值得修、量化分析、SLI/成功率。

igoingdown Bundle

查询 new-api / one-api 系 LLM 网关的当日用量（按模型的花费/tokens/调用次数 + 账户余额）并通过 lark-cli 飞书私聊推送。当用户询问"今天 LLM 用了多少"、"new-api 用量"、"模型花费"、"token 消耗"，或要求配置/排查每小时用量推送定时任务时使用。

igoingdown Bundle

拉飞书 Bug 群的新消息(含最近 N 天历史)按新→旧倒序、message_id 去重挑一条未处理的 bug 进行判定/定位/生成分析报告,展示给用户 review。严格串行:未 review 完的不推进。依赖 lark-event/lark-im/lark-contact 通用 skill + 项目侧 oncall/debug/code-analyze skill(或直接调 aliyun-sls/bytebase/signoz MCP)。当用户在 `/loop 5m /bug-triage-loop` 中调用时触发,或用户手动运行 `/bug-triage-loop` 时触发。

igoingdown Bundle

经 tool-bridge 网关的 mcp/meego 查询飞书项目(Meego)工作项——按人/按时间/按角色筛需求和缺陷、查工作项状态与详情、加评论。使用时：用户想查"我本周要关注的需求/bug"、"某人参与的工作项"、"这个 bug 什么状态"、按 MQL 搜工作项、给工作项加评论时触发。依赖 tool-bridge SK（secrets.sh 注入），空间配置见 secrets.example.sh。

igoingdown Bundle

带宠物家庭自驾旅行规划。给定起点、终点、旅行天数和假期类型， 生成完整的宠物友好旅行方案（自驾+火车），包含避峰日期推荐、 充电计划（理想4C优先，参考信息请通过理想APP确认）、 区域住宿推荐和景点推荐。 当用户说"帮我规划XX到XX的旅行"、"五一出行方案"、 "带狗自驾"、"假期旅行规划"、"家庭旅行"等时触发此 Skill。 车型为 Tesla Model Y，充电优先选择理想4C超充站。

jamillazarev Skill

Offboard an agent or a person — what they own and block surfaces first, then reassignment, then the archive.

jamillazarev Skill

Grow the team — an agent built by the role-builder, or a real person by invitation.

jonathaneciana17-maker Skill

Master comprehensive evaluation strategies for LLM applications, from automated metrics to human evaluation and A/B testing.