AI & ML

misssophiesterling Skill

Turn a workflow that just worked into a reusable Agent Skill (a SKILL.md folder). Use when the user says "turn this into a skill", "save this workflow" or "make this repeatable", or when the same multi-step job has now been done twice.

misssophiesterling Bundle

Full playbook for making an AI coding agent's plan usage last longer on any model. Covers pinpointed prompts, budgets, lean context, cheap subagents, quiet tools, short answers, progress files, skills, cost experiments and window timing. Use at the start of heavy agent sessions, or when the user hits or worries about Claude Code, Codex or Gemini CLI usage limits.

misssophiesterling Skill

Measure and trim what loads before the first message (AGENTS.md, CLAUDE.md or GEMINI.md, memory, installed skills, MCP servers, plugins). Use when context is already full at session start, limits run out faster than expected, the user asks to clean up instruction files, skills or MCP servers, or as a monthly checkup.

misssophiesterling Skill

Plan and review on the strong model while cheaper models do narrow jobs in subagents, under a spawn cap. Use for large or parallel work (bulk edits, many searches, triaging batches of emails or tickets, drafting to a template), or when the user says "use subagents", "delegate" or "use a cheaper model".

misssophiesterling Skill

Schedule one tiny, cheap prompt before the workday so a Claude or ChatGPT/Codex plan's 5-hour usage window starts early and resets sooner. Use when the user runs out of the 5-hour limit partway through the morning, or asks to warm up, ping or pre-start their usage window on a schedule.

qualcomm-linux Bundle

Print a concise summary of the Qualcomm Linux device this runs on: board model and SoC (device tree + socinfo), OS release including BUILD_ID, kernel, uptime and systemd state. On request, produce a full release report instead: firmware inventory from qcom_socinfo, UEFI version from the EFI loader variables, Yocto layer revisions from /etc/buildinfo, consistency findings and a release fingerprint — the QLI 2.x Ver_Info.txt replacement. Use when asked to "print device info", "what board/SoC is this", "which image/build is this board running", "release report", "what firmware is on this board", "which meta-qcom commit built this image", "reconstruct Ver_Info", or to capture a baseline before tests. Read-only; runs on the booted target, or offline on a capture someone pasted. Do NOT use for health diagnostics or host-side build questions (see qcom-yocto-build-image). This is an example skill and the authoring template for this catalog.

qualcomm-linux Bundle

Validate that a Qualcomm Linux board boots to a working login shell by watching its serial console: wait for the login prompt, log in with the image's default credentials, and check uname, /etc/os-release (incl. BUILD_ID) and systemd state, reporting BOOT-VALIDATION: PASS/FAIL. Use after flashing, when asked to "validate the board boots", "check the board over serial", "did the new image come up", or "watch the boot console". Needs the serial console attached to this host (or reachable over ssh). Do NOT use for flashing (see qcom-flash-qdl) or for diagnosing a healthy-but-degraded system.

mina-atef-00 Bundle

Connect Hermes Agent to Zed via the Agent Client Protocol.

mina-atef-00 Skill

Expert knowledge of Starship cross-shell prompt — installation, CLI, TOML config, all modules, style strings, presets, advanced features including transient prompts, custom modules, palettes, right prompt, and Fish shell integration.

mina-atef-00 Bundle

Wire any OpenAI-compatible LLM API into Hermes Agent as a custom provider — Cloudflare Workers AI, self-hosted vLLM, Ollama, LM Studio, or any service that implements /v1/chat/completions.

scagogogo Skill

One-time setup for versions-skills plugin, SDK, CLI, and MCP server. Run first before using any version-* skill.

scagogogo Skill

Invoke version operations via the versions MCP server — tool naming, parameter conventions, server configuration, and batch patterns.

scagogogo Skill

Parse, validate, and extract structured components from version strings via SDK, CLI, or MCP. Covers NewVersion/MustParse, all Is* type checks, Segments, Core, Clone, SuffixWeight, custom parser options.

vsevolodbazhan Skill

Request cross code review from other model and/or harness.

ivelten Skill

Model errors as Haskell domain types with structured context, hand-written Show for humans, explicit-tag ToJSON for logs, and conscious choices about Exception/FromJSON. Use when designing error ADTs, deciding between Either/Error effect/throwIO, writing instances for errors, or reviewing error-handling code.

guysilvere Bundle

Administration des outils IA de l'écosystème Agence Bulles — OpenCode ET Antigravity (Google). Création/validation de SKILL.md, gestion des fichiers agent.md (agents OpenCode + subagents Antigravity), permissions, config (opencode.jsonc, GEMINI.md, workflows, commandes), plugins et sécurisation des tokens. Charger dès qu'une demande concerne la configuration d'un outil, un agent, une skill, une permission, une commande ou un secret en clair.

pixee Skill

Show Pixee platform health, license status, and AI model availability for the currently connected deployment.

guysilvere Bundle

Conventions Git/GitHub Agence Bulles — branche testing unique, commits Conventional Commits, versioning SemVer, notes de release et descriptions de PR, merge et pull, MCP GitHub. À utiliser pour toute opération de versioning (via ops-quality).

mypropper Bundle

Foundation for every Propper agreement task — authenticate, route to the right tool, read the agreement status model, and triage scope and permission errors. Load this before creating, sending, tracking, voiding or importing anything in Propper, and whenever a Propper tool returns an authentication, scope or state error.

roboco-io Bundle

Karpathy LLM Wiki 패턴 구현. 프로젝트 문서(스펙/ADR/회의록)를 wiki/raw/에 모으고 LLM이 컴파일하여 wiki/에 교차참조 마크다운 위키를 생성·갱신. qmd 설치 시 하이브리드 검색 자동 활성, 미설치 시 INDEX.md 라우팅으로 graceful degrade. LanceDB 선택적 벡터 인덱스. Obsidian 호환. 하위 명령 - init, ingest, query, lint, sync, export, qmd-index, lancedb-sync.

ftrout Bundle

Prioritise vulnerabilities and produce a ranked remediation plan using CVSS v3.1/v4.0, EPSS, CISA KEV, exploit maturity, asset exposure and criticality, compensating controls, and business context, with tiers (P1..P4), SLAs, risk statements, and asset-owner communications. Use it whenever someone shares a scanner export, a CVE list, a vendor advisory, a "should we patch this now" question, a KEV notification, or a Patch Tuesday summary and wants to know what to fix first, how urgent a specific CVE is for their environment, what to do when a system cannot be patched, how to write a risk acceptance or exception, or how to explain a vulnerability decision to an asset owner or executive. Also use it to tune a scoring model or SLA policy.

ftrout Bundle

Turn raw threat intelligence (vendor reports, ISAC bulletins, CISA/CERT advisories, STIX bundles, MISP events, social-media threads, leak-site posts, dark-web chatter) into actionable products: extract TTPs and indicators, grade the source with the Admiralty code, assess relevance against the organisation's threat profile and PIRs, apply the Diamond Model, and write flash alerts, actor profiles, and weekly digests with TLP markings plus concrete detection and hunt asks. Use it whenever someone pastes or links a threat report and asks "is this relevant to us", "what should we do about this", "summarise this for the SOC / leadership", "write up this actor", "what is in this STIX bundle", "build our threat profile", "draft this week's intel digest", or wants to know whether an intel claim is credible.

mannyrodriguez-wk Skill

Live-demos Workiva's GRC (governance/risk/compliance) MCP surface to a prospect: control environment overview, risk-and-control-matrix (RCM) review, risk-based audit planning, remediation/issue aging, and policy search. Use when the user says: '/grc', 'show me the GRC demo', 'show me the audit/compliance demo', 'demo controls and risk to this prospect', 'demo the RCM', 'demo risk-based audit planning', or 'show me the compliance heatmap'.

mannyrodriguez-wk Skill

Live-demos Workiva's Documents/Spreadsheets MCP surface for Financial Reporting scenarios: filing review, accounting-standard (ASC) checks, period-over-period filing comparison, feature/link audits at scale, data-lineage tracing, and the 'Auto-SOI' style raw-dataset-to-styled-document workflow. Use when the user says: '/reporting', 'show me the financial reporting demo', 'demo document review to this prospect', 'demo Auto-SOI', 'review this filing for accounting changes', or 'compare these two filing periods'.

mannyrodriguez-wk Skill

Live-demo Workiva's MCP surface for sustainability/ESG reporting scenarios — cross-workspace CSRD/ISSB regulatory scoping, joining ESG program data with financial/subsidiary data, and producing a scoping matrix plus executive summary. Trigger phrases: '/sustainability', 'show me the ESG demo', 'show me the ESG/CSRD demo', 'demo sustainability reporting', 'demo CSRD/ISSB scoping', 'scope this prospect for CSRD/ISSB'.

hiyabh Bundle

Search the CodeGuilds community registry (3,000+ skills, agents, MCP servers, hooks, prompts for Claude Code) for an existing tool that does what the current task needs, then install it — only after the user approves. Use PROACTIVELY whenever Claude is about to tell the user it lacks a capability, tool, skill, agent, MCP server, or integration to complete a task: before giving up, building from scratch, or saying "I can't do X", first search CodeGuilds for something that already does it. Also use when the user explicitly says: "codeguilds", "find a tool/skill/MCP for", "is there a skill/agent/plugin for", "search the registry", "תחפש כלי", "יש תוסף/סקיל ל...", "תתקין כלי בשביל", "חפש ב-codeguilds", "יש משהו מוכן ש...". NOT for tasks Claude can already do with built-in tools or existing skills — only when a genuine capability gap exists.

jtydhr88 Skill

The tone-melody conflict detector and how to read it (倒字检测器). The one objective, computable criterion this library has - Mandarin dao-zi, where a syllable is heard as a different word because the melody enters it against its tone contour. Covers the correct model (incoming interval versus the syllable's own contour, not previous-end versus next-start), how to run it, how to read a risk score, the three boundaries that make it a risk indicator rather than a verdict, and the five ways to fix a flagged syllable. Use when fitting Mandarin words to a melody, when a sung line is misheard as different words, when checking a draft before delivery, or when a melody must be adjusted to fit words already written. 倒字、依字行腔、字调、声调、检测器、唱出来听成了别的字、唱成别的字、字唱错了、tone_check。

jtydhr88 Bundle

The case layer - how to take a finished lyric apart and what three working lyricists say about their own habits (案例与拆解). A fixed twelve-row worksheet for dissecting any song, keyed to the library's own fields so every finding lands in a skill instead of staying an impression; the rule that a real lyric is read from the middle outward, not from the theme down; what Yao Qian, Zhang Li and Hong Yuan each say about opening lines, restraint, revision, rhyme width, colloquial versus literary diction and how much to leave the composer, plus the four places they openly disagree and how to pick a side; worked dissections of songs the sources analyse in full, including one song two of them discuss independently; and the failure patterns they name - flat, scattered, overlong, bland, blunt, bare, stuffed. Use when studying a lyric you admire, when you need a real example of a technique, when a draft feels wrong and you want a model to compare against, or when asked to critique a song. 案例、拆解、词评、作品分析、词人自述、改稿对照。

jtydhr88 Skill

Compiling a finished lyric into generation-backend input (词稿 → 后端). Section tags and how they map, why line breaks inside a section are read as phrase boundaries rather than layout, what the backend does with punctuation and parentheses, the division of labour with the arrangement library's compiler, the backends that accept lyrics at all and what each does with them, and the checks that catch a lyric which will be sung wrong. Use when handing a finished lyric to Suno, YuE2 or MiniMax, when a generated vocal phrases the words wrongly, when sections come out in the wrong order, or when deciding what belongs in the lyric field versus the style prompt. 歌词、Suno、YuE2、段落标签、编译、交给模型、断句。

mannyrodriguez-wk Skill

Analyzes each security/compliance question in a customer or prospect RFP or security questionnaire against Workiva's documented internal security posture (via the internal Workiva GRC/Documents MCP, read-only), drafts a proposed answer for each question, and explicitly flags any question where Workiva's documented posture doesn't clearly cover the ask. Use when the user says: 'analyze this RFP's security section', 'answer these security questionnaire items', 'check our posture against [question]', 'draft responses to this security questionnaire', or 'does Workiva's documented posture cover this RFP question'.

clearailhc Skill

Use when working inside the ClearAI preset and you need the loop's contract rather than its prompt — how a claim becomes a fact (hypothesis → verification → observation → evaluation → evidence → fact), which tool is the only completion verb, what observation admission does and does not decide, who may write a verdict at each level (L0–L4), what "doer does not judge themselves" costs you, and what happens when a step fails admission. Load it before planning multi-step work, before declaring anything finished, or when a step keeps getting blocked.

shreyp087 Bundle

Evaluate third-party libraries, starter repositories, agent skills, and plugins for a specific project bottleneck; compare reuse against a small implementation and inspect the chosen dependency before adoption. Use for dependency scouting, not ordinary coding with an established stack.

shreyp087 Bundle

Split a hackathon build into independent agent or teammate tasks with explicit interfaces, file ownership, integration order, and bounded verification.

ficilcom Bundle

Work out the order and the loose ends before turning down an offer or withdrawing from a selection process: whether the offer being kept is actually settled in writing, whether each decline is mid-selection, post-offer, or after acceptance, who besides the employer has to be told when the application came through an agent, a school recommendation, or a referral, which reply deadlines have passed, and what documents, expenses, or borrowed items are still outstanding. Use when a user holds more than one offer and needs to decline the rest, wants to withdraw from a process they are no longer pursuing, or has been asked to turn down other employers as a condition; do not use to decide whether to decline, to rule on what declining after acceptance means legally, or to send any message.

stifleur390 Skill

Intégration AI/LLM dans Laravel 13 avec le Laravel AI SDK. Agents, tools, embeddings, RAG, vector stores, streaming. Utiliser quand on intègre de l'IA dans une app Laravel, on crée des agents, ou on fait du RAG.

stifleur390 Skill

Serveur MCP (Model Context Protocol) pour Laravel. Exposer des données et actions via le protocole MCP pour les outils AI. Utiliser quand on crée un serveur MCP, on expose des données à un assistant AI, ou on configure Laravel Boost.