Context & AI engineering
25,864 rows, by stacks then stars
1,865 of these skill files have been read, by 1,328 distinct authors, and what they tell an agent counted. What Context & AI engineering authors agree on, and what they forbid
alexclymo/claude-tracker-skills/skills/tracker-audit Skill
0★ repoUse when the user wants to audit the task tracker, clean up the task docs, or says the docs feel stale, out of date, or contradictory — triggers like "audit the task tracker", "clean up the task docs", "the docs feel stale", "reconcile the tracker", "health check the tracker", "do the tracker docs still match reality". Sweeps the whole repo — tracker/, CLAUDE.md, architecture and navigation docs, and code comments — for stale and contradictory claims, reports them ranked by how likely each is to cause a wrong agent decision, and fixes them only on explicit approval. Manages project task *documents* on disk, not Claude Code's in-session task tools (TaskCreate/TaskList) — those are scoped to one conversation and never persist; this skill audits the durable, git-tracked record instead.
murphy977/3d-eyes/skill Skill
0★ repoUse when the user reports a 3D model or scene looking wrong (black/invisible/gigantic/untextured model, holes, weird shading), asks to check or QA a 3D file (GLB, GLTF, OBJ, PLY, STL, .splat), works with Gaussian splats, or generates 3D assets with AI and wants them verified. Requires the 3d-eyes MCP server.
yuanGao0816/prompt-rail Skill
no license0★Optimize prompts with a measured train/test dual-gate loop that resists memorizing eval cases. Use when the user wants to write or improve a prompt, tune a system prompt, build prompt evals with holdout, prevent prompt overfitting/背题, run prompt-rail, or iterate intent-classification prompts with separate train and test suites. Fork of prompt-smith plus anti-overfit rails.
shihchengwei-lab/session-analytics Skill
0★Weekly self-improvement loop for AI coding agents, plus ad-hoc usage analytics. Each agent analyzes its OWN tool's local session logs over the last 7 days (Claude Code via /insights artifacts, Codex via ~/.codex/sessions, others via their logs), reports success rates, friction patterns, stalled sessions, tool/token usage — and in a weekly review proposes evidence-tied workflow improvements: a small capped rules block in the tool's config (diff shown, user approves, block self-refactors instead of accumulating), mechanical harness fixes (hooks/env/permissions) preferred over prose rules, and skill/config hygiene both directions (hot paths reinforced, installed-but-never-invoked skills flagged for disabling; rule evidence re-validated across model changes). Trigger on "weekly review", "improve my workflow", "分析我的 session/使用紀錄", "which sessions failed", "my success rate", "compare my experiment runs", "哪些 skill 沒在用/停用沒用的 skill", or bare invocation (= 7-day overview). Reads local data only. Not for regenerating the official /insights report, or deep-reading one full transcript (expensive — see Cost rules).
proluct/skill-press/skills/skill-press-polish Skill
0★ repoDiagnose and improve an EXISTING Claude Code skill with a diagnose, fix, re-diagnose pass, whether skill-press built it or not. Use when the user says "polish this skill", "improve my X skill", "this skill keeps misfiring or confusing the agent", "clean up this SKILL.md", or after a skill-press run offers a polish pass. Not for creating new skills (use skill-press) and not for plugin-managed skills, which cannot be edited in place.
CodingCossack/writing-claude-design-prompts Skill
0★Use when turning product notes, briefs, specifications, or mixed mission packs into prompts for Claude Opus, Fable, or another design-capable Claude model whose primary result is a visual artefact, frontend, redesign, prototype, design system, or image-to-code implementation.
sonu07-star/agent-handoff-skill/skills/agent-handoff Skill
0★ repoCreate, validate, and resume structured handoff documents that let another AI agent continue work without hidden conversation context. Use when pausing or transferring unfinished work, approaching a context limit, switching agents or sessions, reporting a blocker, recovering after interruption, resuming from HANDOFF.md, or auditing a handoff for evidence, actionable next steps, secret leakage, and unsupported completion claims.
stone16/stometa-skills/skills/doc-steward Skill
0★ repoSteward agent-facing repository documentation against a tiered house standard. Use when auditing or grading AGENTS.md, CLAUDE.md, SKILL.md, .claude/rules, ADRs, or DESIGN.md for structure, drift, frontmatter, or broken routing; explicitly previewing or applying gated low-risk fixes after an audit; or explicitly capturing one documentation lesson. Not for prose copy-editing, application-code review, product-interface design, or generic formatting.
SubliminalCoding/build-it-out/skills/build-it-out Skill
0★ repoRun a hardened coordinator-worker-reviewer build board for substantial implementation, hardening, migration, or decompose-build-review-iterate work. Use when a user says "build it out", "use the workflow", "run the board/swarm", requests a hardening pass, or wants independent orchestration and review. Uses a read-mostly Claude coordinator, exact-pane Codex workers, deterministic Git/scope/test gates, and a fresh capability-restricted Claude reviewer with bounded repairs. Do not use for a quick conversational answer or a trivial single-file edit.
Optimize claude fable 5 prompts
IchenDEV/claude-fable-5-prompt-optimizer/plugins/claude-fable-5-prompt-optimizer/skills/optimize-claude-fable-5-prompts Skill
0★ repoClarify, audit, and rewrite rough or existing prompts for Claude Fable 5 using Anthropic's official prompting guidance. Use when a user asks to optimize, improve, migrate, debug, or design a prompt for Claude Fable 5, including requests phrased as Fable or Fabel prompt optimization, Fable/Fabel 提示词优化, or when an underspecified idea needs guided clarification before a copy-ready prompt. Ask only for missing information that would materially change the result; otherwise preserve intent and produce a lean, context-rich prompt with explicit outcomes, boundaries, evidence, output requirements, and completion criteria.
The same file is in 1 other repository
YMIR777/vibecoding-prompt-workflow/skills/prompt-first-workflow Skill
no license0★ repoApply first-principles task framing with a bounded read-only preflight before substantive work. Use for projects, changes, debugging, reviews, research, writing, design, and other tasks where goals, context, constraints, approval boundaries, acceptance criteria, or output requirements need to be explicit. Show an executable task definition before substantive work on complex, open-ended, state-changing, or decision-bearing requests; do not force ceremonial rewrites for simple, already-clear answers or low-risk actions.
Chulinuwu/skillloom/skills/autonomous-learning Skill
0★ repoReview completed Claude Code, Codex, or compatible agent work for durable declarative knowledge and reusable procedural knowledge. In Skillloom Hermes mode, curate one bounded second-brain outcome only after a checkpointed meaningful delta and policy-promote only validated safe skill candidates. Use when explicitly invoked, when a Stop hook requests learning review, when an invoked host completes non-trivial work, or when repeated corrections reveal a stable workflow.
neilchen2000-pixel/agent-skill-stack/skills/agent-skill-stack Skill
0★ repoFind, evaluate, assemble, and runtime-activate the smallest compatible set of AI Agent Skills for an end-to-end natural-language goal. Use as the primary orchestrator when a request spans multiple replaceable capabilities or asks for a sustained, complete, or end-to-end outcome—even when the user never says “Skill”—and use domain Skills only as downstream specialists. Also use for multi-Skill selection, installed-Skill audits, conflict checks, low recall, indirect helpers such as humanizers or compliance checks, and project routing with dependency closure. Search local Skills, registries, GitHub, and OpenCLI; compare adoption, verified fit, safety, and overlap; block incomplete handoffs. Do not use for a single bounded action or locating one known/common Skill; use that domain Skill or the generic find-skills workflow.
HustleCoding/clear-agent-communication/skill/clear-agent-communication Skill
0★ repoExplain an AI agent's multi-step work clearly, selectively, and in plain language. Use when investigating problems, changing files or data, running tools, making consequential decisions, encountering blockers or uncertainty, or reporting verification and completion. Communicate observable actions and evidence without exposing private chain-of-thought or narrating every operation.
Acelogic/sharpemu-agentic-toolkit/skills/sharpemu-nid-swarm Skill
0★ repoRun evidence-backed SharpEmu PS5 NID work from unresolved-import traces through KawaiiDRA reverse engineering, subsystem clustering, isolated worktree implementation, integration, and multi-game regression. Use when an AI coding agent is asked to triage, identify, reverse engineer, implement, batch, delegate, or regression-test SharpEmu NIDs; compare NID coverage with another emulator or PR; or build a parallel NID work queue from logs, CSV catalogs, Aerolib, firmware, Kyty, or game traces.
arkaigrowth/agent-skills/caveman-compress Skill
0★ repoCompress natural language memory files (CLAUDE.md, todos, preferences) into caveman format to save input tokens. Preserves all technical substance, code, URLs, and structure. Compressed version overwrites the original file. Human-readable backup saved as FILE.original.md. Trigger: /caveman:compress <filepath> or "compress memory file"
wu1chenghui/hermes-math-template/skills/agent-memory-architecture Skill
no license0★ repoDesign persistent, tiered memory systems for LLM agents — abstraction-compression pipelines, selective forgetting, hierarchical retrieval, and container-persistence engineering.
LiuHHHYuuu/agent-skills/skills/harness-engineering Skill
0★ repoUse whenever the user wants to design, improve, or audit agent-first engineering systems, self-improving Codex workflows, autonomous PR/review loops, vibe-coding guardrails, new systematic projects, greenfield project scaffolding, project knowledge bases, golden principles, cleanup agents, architecture constraints, readable code style, function comments/docstrings, eval loops, durable logs, human judgment allocation, or human approval boundaries. Trigger on phrases like harness engineering, self-improving agents, agent-first, vibe coding, new project, start a serious project, build a system, AGENTS.md, golden principles, drift, garbage collection, cleanup agent, context management, code style for agents, reviewable code, agent-to-agent review, or Building self-improving tax agents with Codex.
wakamiya0315/delegate-to-claude/skill/delegate-to-claude Skill
0★ repoUse this skill by default before the supervising agent directly edits repository code when work adds or updates tests, likely changes two or more files, creates a non-trivial module, investigates a failure, performs a focused refactor or code review, or likely needs more than three repository tool calls. Delegate the bounded execution slice to a supervised Claude Code Sonnet worker, using a concise prompt when possible. Direct editing is only for an unambiguous low-risk change of about ten lines or less in one file with no tests or investigation. Keep architecture, ambiguous product decisions, security-sensitive work, secrets, external side effects, commit/push/deploy, and final approval with the supervisor.
p404/monodraw-llm/.claude/skills/monodraw-llm Skill
0★ repoTurn natural-language descriptions, ASCII/text diagrams, screenshots, and existing `.monopic` or `.monojson` files into native editable Monodraw documents. Create, reconstruct, inspect, validate, and regression-test complex diagrams with separate resizable frames, nested groups, attached or free-endpoint connectors, arrows, shadows, z-order, and pinned connector labels. Use whenever an LLM should sketch a diagram in ASCII and export it to Monodraw, recreate a screenshot or legacy v3/current v5 document 1:1, or diagnose Monodraw serialization and editability.
craxelfn/claude-to-codex-migrator/skills/claude-to-codex-migrator Skill
no license0★ repoMigrate Claude skills, plugins, assistant packages, ZIP archives, pasted files, repository trees, or implementation documentation into clean Codex Skills or plugins. Use when Codex must inventory a source package, choose Skill versus plugin architecture, preserve behavior, rewrite platform-specific instructions, validate MCP or hooks, remove Claude branding from the distributable output, and produce migration and cleanup reports.
patrikherak/llm-privacy Skill
0★Reversible PII tokenization for working with sensitive data. Masks personal data (emails, phones, IBANs, cards, national IDs, tagged names/addresses) behind stable placeholder tokens BEFORE text enters model context, and restores the real values in the final output — deterministic code, no ML, no network. Use when reading, querying, or processing data that contains (or may contain) personal information of real people — customer records, database query results, CRM exports, support tickets, log files — so the model never sees the actual PII but the user still gets a complete answer.
lniass/social-skills/skills/social-agent-public-workflows Skill
0★ repoWrite, review, approve, and schedule Facebook posts for a business through the hosted Social Agent service. Use this whenever someone asks for a social media post or caption, wants to see posts already prepared for them, wants to approve or schedule one, wants to connect their Facebook Page, or wants recurring posting set up. Post copy comes from the hosted service and is never written locally. Covers first-time setup through secure Handled verification as well as returning users who already have a project.
flamefalcon7/youtube-transcript Skill
0★Transcribe YouTube videos locally with Whisper into agent-ready transcripts — no captions, no cloud. Per-channel/domain vocabulary memory makes it more accurate every time the user corrects a term. Use when given a YouTube URL to transcribe, process, summarize, or sync into a knowledge base.
TJCurnutte/cohort-method/.agents/skills/cohort-method Skill
0★ repoRun multi-agent cohort reviews, creative panels, UX swarms, and cross-QC passes (e.g. "10 agents / 5 cohorts"). Spawns parallel persona or scoped agents, synthesizes consensus, and soft-merges findings into actionable next ships without reshuffling live work. Use when the user asks for a cohort review, creative panel, UX swarm, taste panel, multi-agent QC, or phrases like "spin up N agents" / "N agents / M cohorts".
wretcher207/dead-pixel-skills/skills/codex-prompting Skill
0★ repoUse this skill when the user wants an agent to create, rewrite, critique, or harden a paste-ready prompt for OpenAI Codex from a rough request, bug report, screenshot, specification, or repository task. Covers implementation, diagnosis, review, UI work, refactors, and long-running execution. Do not use it merely because the current agent is Codex or because the user wants the target task executed.
h8nc4y/multi-agent-delegation Skill
0★Prevent no-op and silent delegation failures when an orchestrator agent hands work to subagents. Use when a subagent re-delegates to a child agent and idles instead of working, when a completion notice arrives but no artifacts exist (git status shows no changes), when execution-environment differences (WSL / Git Bash / PowerShell / sandbox) fake success silently, or before writing any delegation prompt, especially when multiple writers could share one checkout containing WIP. Canonical source for mandatory delegation-prompt clauses, artifact verification of completion notices, resume-based recovery, and ledger-as-spec parallel delegation. Symptom keywords: silent failure, no-op delegation, "delegated and waiting", 完了通知なのに成果物が無い, 子へ再委譲して待機, 空振り, 無言失敗.
avnath13/evalpilot Skill
0★Run the full agent-eval quality flywheel on autopilot from a goal in evalpilot.config.yaml, prepare an eval set from traces or synthetic scenarios, run inference, grade with autoraters and custom rubrics, cluster the failures, ship a targeted fix, and loop until quality targets are met. Use whenever the user wants to evaluate, benchmark, find quality bugs in, or improve an AI agent, chatbot, or LLM app, including "eval my agent", "why is my agent wrong", or "prove this fix worked".
efaimo-ai/efaimo Skill
0★Audit MCP servers and Agent Skills for quality and context-window cost. Use when the user wants to check an MCP server's token cost, lint a skill, verify 2026-07-28 MCP spec readiness, or add an efaimo badge or CI gate.
MaikoCode/brainrotshorts-agent-skills/skills/brainrotshorts Skill
no license0★ repoCreate, edit, render, and retrieve vertical short-form videos with BrainrotShorts through its remote MCP server or REST API. Use when an agent needs to make AI brainrot dialogue videos, Reddit story videos, AI voice-over videos, scripts, scenes, captions, backgrounds, timed overlays, project automations, or manage existing BrainrotShorts projects and renders.
pecigonzalo/agent-skills/skills/mcp Skill
0★ repoDiscover, configure, and call MCP (Model Context Protocol) servers. Use this when interacting with MCP tools or resources in Pi, inspecting schemas, authenticating servers, trying ad-hoc endpoints, or generating typed MCP clients and CLIs.
Paldom/terminaltor/.claude/skills/add-skill Skill
0★ repoAuthors or improves an Agent Skill in this repo via the eval-first workflow - scope one purpose, write trigger evals, draft SKILL.md, validate, iterate. Use when the user asks to create, add, write, refine, or fix a skill ("add a skill for X", "improve triggering of Y"). Not for scaffolding a new repository or editing infrastructure (hooks, CI, the validator/scripts, docs).
The same file is in 11 other repositories
SubliminalCoding/orchestrate-it-out/skills/orchestrate-it-out Skill
0★ repoRun a high-level GPT-5.6 Sol orchestration workflow for substantial coding work. Sol inspects a clean Git repository and produces a comprehensive sealed plan with exactly five bounded workstreams; the controller then spawns exactly five Claude Fable 5 workers, dispatches each brief to an exact pane, waits by dependency wave, requires receipts, and runs mechanically enforced scope and verification gates. Use when the user asks for Sol to orchestrate five Fables, wants a thorough plan followed by a five-agent build, or explicitly says "orchestrate it out." Do not use for small edits, non-coding requests, dirty repositories, or requests that require independent semantic review.
0xkaveh/debug-orchestrator/skill/debug-orchestrator Skill
0★ repoOrchestrate evidence-led, security-first debugging across local code, tests, runtimes, browsers, and CI. Use when Codex must diagnose a bug, failing test or build, runtime exception, intermittent behavior, frontend/browser failure, performance regression, or failing GitHub Actions check; reproduce the problem before editing, maintain at most three falsifiable hypotheses, route to the cheapest useful tool, apply a minimal root-cause fix, add a regression test, and verify the result with recorded evidence.
paladini/agent-skills/skills/devto-publish-expert Skill
0★ repoResearch, write, optimize, validate, and publish specialist technical articles for DEV Community with exactly four relevant tags and a draft-first safety workflow. Use when creating, rewriting, reviewing, scheduling, or publishing DEV.to tutorials, project launches, engineering explainers, case studies, or opinionated developer posts, including through a DEV/Forem MCP server or API.
saezbaldo/deploytoagents-skills/deploy-app-to-agents Skill
0★ repoAudit and distribute an existing app, API, SaaS, MCP server, or agent so AI assistants can find, verify, recommend, and invoke it. Use when a user asks how to launch or distribute software to agents or LLMs, improve agent discoverability, publish MCP/A2A metadata, create machine-readable discovery surfaces, compare distribution channels, or measure whether unbranded prompts mention their product.
Implementation handoff execution
Pingbook-co/agent-skills/skills/implementation-handoff-execution Skill
0★ repoUse when an approved phase and handoff documentation set needs serial or controlled-parallel implementation, continuation, resumption, context recovery, validation, integration, or completion.
xiewxin/agent-skill-maintainer/skills/agent-skill-maintainer Skill
0★ repoUse when reviewing an Agent Skill after real use, user correction, external feedback, or observable decision failures, especially before proposing changes to a Skill repository.
arnaudjnn/billing-tools Skill
0★Add drop-in auth and billing to a Stripe + WorkOS app with the @arnaudjnn/billing-tools library (Next.js, Hono, Bun, Deno, any Node host). Use this skill whenever the user wants to monetize an app, MCP server, or API, charge per token or per credit, add subscriptions or API-key auth, let AI agents self-register with auth.md and pay per request with Stripe MPP machine payments, or wire a token wallet, Stripe Checkout, auto-reload, a Customer Portal, invoices, seat limits, or per-seat token grants, and expose it all as MCP tools, a REST API, or a CLI. Covers the pluggable WorkOS-org adapter (WorkOS-only or WorkOS plus a DB mirror), declarative plans auto-provisioned in Stripe, and a zero-webhook event-polling sync. Triggers on "add billing", "monetize my agent or tools or API", "Stripe and WorkOS", "charge per token", "agent auth.md", "machine payments", or "MPP", even when the library is not named.
RayJason/rayjason-skills/harness-agents-md Skill
0★ repo创建、审计和优化全局与项目级工程化的 AGENTS.md,覆盖作用域、协作、Git、验证与交付。
Gazer2020/zglii/feishu-hook Skill
0★ repoSend Markdown card messages to Feishu (Lark) group chats through a custom bot webhook, with optional signature verification. Use when the user asks to post a notification, alert, status update, or other message to a Feishu/Lark group.
Evidence first durable task orchestrator
x2724009/evidence-first-durable-task-orchestrator/skills/evidence-first-durable-task-orchestrator Skill
0★ repoEnforce a design gate before implementation for genuinely complex or high-risk work. Use only after the task passes the entry gate: explicit durable tracking/cross-session recovery/high-risk public change, or at least two dependency, verification, ambiguity, recovery, or coordination signals. The Agent must define scope, risks, decisions, and acceptance before implementation; evidence records preserve that decision. Do not use for ordinary multi-step work, routine research, testing, or small UI changes alone.
qmayxin/skill-matcher/skills/skill-matcher Skill
0★ repoUse this skill to find, verify, compare, install, or migrate Agent Skills across Codex, Claude Code, Cursor, OpenClaw, OpenCode, GitHub Copilot, WorkBuddy and other SKILL.md hosts. Trigger when users ask “find a skill for X”, “what skills are installed”, “search GitHub deeply”, “is this skill compatible or safe”, or “install/migrate this skill”. 支持“找个技能”“盘点本机 Skills”“全网或 GitHub 深挖”“检查兼容性、安全性”“安装或迁移技能”等中文请求。Discover and evaluate skills only; do not perform the target business task.
blakebauman/skillist-validator/skills/skill-description-tuning Skill
0★ repoDiagnose and fix why an agent skill triggers on the wrong prompts, or fails to trigger on the right ones, by rewriting its description and measuring the change against labelled eval queries. Use when a skill never activates, fires when it should not, activates only if the user names it explicitly, or when the user asks to improve, test, or optimize a skill's description or triggering. Covers writing the description, designing should-trigger and should-not-trigger queries, running trigger-rate evals with a train and validation split, and choosing the winning version.
sapochat/working-with-an-agent Skill
0★Use when designing or sharing a durable human-agent operating agreement: collaboration posture, memory and skill boundaries, correction capture, autonomy budgets, sanitization, and verification standards.
entalpa/entalpa-mcp/plugins/entalpa/skills/entalpa-implement Skill
0★ repoUse this skill when working with Entalpa through its MCP server to implement stories and requirements that were authored on the Entalpa website, inspect the project/stakeholder/story/requirement/subsystem hierarchy, record code references, implementation status, and test coverage back onto requirements, or otherwise perform requirements-engineering work that should preserve traceability and avoid destructive MCP tool calls.
clearlove7818/waa-agent-delegation Skill
no license0★Govern an actual agent delegation only after the primary agent has decided to create or invoke a subagent. Use to choose exactly one EXECUTION_SUBAGENT, TASK_SPECIALIST_SUBAGENT, or explicitly authorized NAMED_AGENT; check capability, platform, permission, and external-effect boundaries; build a risk-sized TASK-006 packet; define handshake, failure, delivery, and handoff; and support primary-agent verification. Exclude complexity alone, delegation discussion, simple direct work, multi-agent demonstrations, unauthorized named-agent activation, permission expansion, and responsibility transfer.
wildcard/codex-to-claude-coordinator/skills/claude-to-codex-coordinator Skill
0★ repoCoordinate Claude Code-to-Codex delegation through OpenAI's official codex-plugin-cc runtime. Use when Claude should ask Codex to review, investigate, implement, resume, transfer a Claude session, run in the background, report status or results, or cancel work while preserving scope, permissions, model intent, and evidence.
Jorgut/context-quality-suite/skills/eval-harness Skill
0★ repoFormal evaluation framework for Codex sessions implementing eval-driven development (EDD) principles
merttcetn/research-swarm/skills/research-swarm Skill
no license0★ repoParallel research without context bloat—a Codex and Claude Code skill using read-only subagents, isolated reports, and evidence-backed synthesis.
bigblueboo/expert/skills/expert Skill
0★ repoConsult GPT-5.6 Pro through the local `expert` CLI for second opinions on coding tasks. Use when the agent needs a high-quality external review, architecture/debugging help, test strategy feedback, implementation plan critique, or analysis of attached code/docs using explicit local files, directories, globs, or stdin. Especially useful for hard, ambiguous, high-risk, or long-running coding questions where a long blocking consult (up to 6 hours by default) is acceptable.
tomkloevekorn/nock-agent-plugin/plugins/nock/skills/nock-mcp Skill
0★ repoSafely operate a configured NOCK MCP server for support-ticket triage, investigation, creation, updates, internal notes, and customer replies. Use when an agent needs to read or change NOCK tickets through MCP.
cloudh0/lc-rules/skills/lc-rules Skill
0★ repoCreate, audit, or change a workspace's routed rules/ tree and concise agent guidance without overwriting existing project instructions. Use when the user says "lc-rules 스킬 초기화 해줘", asks to initialize or refresh workspace rules, discover conventions across multiple projects, prepare rules for a new frontend/backend project, inherit rules from a reference project, connect AGENTS.md/CLAUDE.md/GEMINI.md to project rules, choose Git or branch conventions, says "브랜치 규칙 전환해줘", or checks an existing lc-rules installation.
CodingCossack/writing-claude-5-prompts Skill
0★Use when writing, rewriting, reviewing, shortening, or handing off prompts, system prompts, agent briefs, skills, CLAUDE.md files, or task specifications targeting Opus 5 or Fable 5 — including when a prompt was tuned for an older model, has grown repetitive or over-scaffolded, or is unclear about scope, autonomy, evidence, output length, or stopping. Not for prompts whose primary result is a visual artefact; writing-claude-design-prompts owns those.
014-code/agent-skill-doctor/skills/skill-doctor Skill
0★ repoStatically audit Agent Skill directories for invalid SKILL.md structure, weak metadata, broken local references, packaging problems, and risky script or instruction patterns. Use when reviewing, validating, publishing, installing, or CI-checking a Codex, Claude Code, or compatible Agent Skill, especially before trusting a third-party Skill.
Pandasocool/compound-learning/skill/compound-learning Skill
0★ repo基于开放 Agent Skills 结构,将播客、视频、访谈、演讲、课程和文章转化为可行动、可复盘的长期学习系统,并保存到用户选择的 Obsidian 或飞书文档。适用于 Codex、Claude Code、WorkBuddy 及其他支持 SKILL.md 的 Agent。用户提供内容链接、要求提炼内容并沉淀到知识库、建立学习地图、选择性制定长期实践计划,或要求周、月、半年、年度学习复盘及芒格式逆向分析时使用。也用于首次介绍功能、选择存储端、初始化学习根目录、固定笔记格式、检测或恢复连接。
stella-dust/autoxeo-marketplace/plugins/autoxeo-creator/skills/autoxeo-creator-orchestrator Skill
0★ repoRoute AutoXEO Creator requests across GEO requirement diagnosis, custom skill/plugin blueprinting, scaffolding, marketplace packaging, and service handoff.
Offwhite-Del/contextlean/plugins/contextlean/skills/optimize-agent-context Skill
0★ repoAudit, benchmark, and safely optimize local AI Agent context, instructions, Skills, Context Packs, plugins, hooks, and cross-vendor configuration. Use when ChatGPT/Codex, Claude Code, or another CLI Agent starts with excessive or stale context, routes Skills poorly, repeats tool work, has conflicting rules, or needs evidence-backed before/after optimization with rollback. Do not use for ordinary code edits, model/provider tuning, or unsupported speed claims.
mike-ai-de/fable-workflow-kit/skills/claude/fable Skill
0★ repoApplies a Fable-style output optimizer for complex coding, writing, research, analysis, planning, and prompt engineering. Use when the user invokes /fable or asks for maximum quality from weaker models, rigorous reasoning, stronger prompts, verification, source discipline, or a structured high-output workflow.
lee101/appnz-agent-skills/skills/appnz-hosting Skill
0★ repoDeploy and operate open-source projects on app.nz. Use when an agent needs to publish a static site or server app, create or validate appnz.yaml, deploy an open model as a scale-to-zero Cog or RunPod worker, launch a cloud coding agent, connect the app.nz MCP server, or call the OpenAI-compatible model gateway with APP_API_KEY.