{
"markdown": "# AI Agents Library\n\n## Production-Ready AI Agent Prompts & Skills\n\n\n\n**28 Custom GPT agents** and **140 agent skills** for ChatGPT, Claude Code, and Codex.\n\n[](https://opensource.org/licenses/MIT)\n[](./custom-gpt)\n[](./frameworks/shared-skills)\n[](CONTRIBUTING.md)\n\n[Quick Start](#quick-start) • [Skills](#agent-skills) • [Custom GPTs](#custom-gpt-agents) • [Contributing](#contributing)\n\n
\n\n---\n\n## What's Inside\n\nAgent skills are portable instruction sets that teach AI agents how to execute domain work\nwell — each is a directory with a `SKILL.md` entry point (YAML frontmatter + routing),\nprogressive-disclosure references, and self-contained scripts and data.\n\nTwo things, usable independently:\n\n- **[`frameworks/shared-skills/`](frameworks/shared-skills/)** — 140 agent skills following the\n [Agent Skills specification](https://agentskills.io/specification). Drop into `~/.claude/skills/`.\n- **[`custom-gpt/`](custom-gpt/)** — 28 Custom GPT system prompts for ChatGPT, written to fit\n the GPT builder's 8000-character Instructions limit.\n\n## Quick Start\n\n```bash\ngit clone https://github.com/vasilyu1983/AI-Agents-public.git\ncd AI-Agents-public\n\n# One skill\nln -s \"$PWD/frameworks/shared-skills/skills/ai-rag\" ~/.claude/skills/ai-rag\n\n# All of them\nfor s in frameworks/shared-skills/skills/*/; do\n ln -s \"$PWD/$s\" ~/.claude/skills/\"$(basename \"$s\")\"\ndone\n```\n\nSymlinks mean edits take effect immediately — no re-sync step.\n\nFor a Custom GPT: open `custom-gpt///01_*.md`, paste into the GPT builder's\nInstructions field, and attach the sibling `02_sources-*.json` as Knowledge if present.\n\n## Agent Skills\n\n| Family | Count | Covers |\n| --- | ---: | --- |\n| `software-*` | 34 | Backend, frontend, iOS, Android, desktop, security, payments, performance |\n| `ai-*` | 21 | RAG, evals, prompt engineering, pretraining, post-training, inference, MLOps |\n| `qa-*` | 17 | Test strategy, debugging, resilience, observability, accessibility, Playwright |\n| `foundations-*` | 16 | Queueing, control, information, game, decision, reliability theory for systems |\n| `ai-coding-agents-*` | 13 | Coding-agent runtime internals: sandboxing, permissions, sessions, tools |\n| `dev-*` | 10 | Git workflow, API design, dependency management, code graphs, planning |\n| `agents-*` | 6 | Hooks, MCP servers, skill authoring, swarm orchestration, repo memory |\n| `data-*` | 5 | Lakehouse, streaming, SQL optimization, analytics engineering |\n| `ops-*` | 4 | DevOps platform, incident response, cost optimization, CI/CD |\n| `document-*` | 4 | docx, pdf, pptx, xlsx |\n| `docs-*` | 3 | Docs-as-code, AI-ready PRDs, note retrieval |\n| `research-*` | 3 | arXiv triage, GitHub mining, research method extraction |\n| `product-*` | 2 | Product management, help centers |\n| `gamedev-*` | 2 | Godot, Roblox |\n\nTwo areas are unusually deep and harder to find elsewhere: **coding-agent runtime internals**\n(how agent CLIs actually implement sandboxing, permissions, and session resume) and\n**systems-theory foundations** (queueing and control theory applied to real design decisions,\nnot as abstract math).\n\nFull breakdown: [`frameworks/shared-skills/README.md`](frameworks/shared-skills/README.md).\n\n### How a skill is structured\n\n```text\n/\n├── SKILL.md # Entry point — frontmatter + routing. Kept small.\n├── references/ # Depth, loaded only when the task needs it\n├── scripts/ # Deterministic helpers — no model call needed\n├── assets/ # Templates and output contracts\n└── data/ # sources.json and curated data\n```\n\n`SKILL.md` stays cheap enough to hold in context; `references/` carry the detail and cost\nnothing until read.\n\n## Custom GPT Agents\n\n28 agents across six categories:\n\n| Category | Count |\n| --- | ---: |\n| Productivity | 8 |\n| Lifestyle | 7 |\n| Programming | 5 |\n| Education | 3 |\n| Research & Analysis | 3 |\n| Writing | 2 |\n\nEach folder holds `01_*.md` (the system prompt) and, where useful, a curated sources JSON.\n\n## Repository Layout\n\n```text\nAI-Agents-public/\n├── custom-gpt/ # 28 Custom GPT prompts by category\n├── frameworks/\n│ └── shared-skills/skills/ # 140 agent skills\n├── CONTRIBUTING.md\n└── LICENSE\n```\n\n## How This Library Is Built\n\nThese skills are the public subset of a larger private library that is run as an\nengineered system, not a prompt collection. Every skill here passed through that\nsystem's quality machinery:\n\n- **A routing engine** — 8 domain routers dispatch 500+ decision scenarios, each an\n ordered skill chain with an explicit gate and a named failure mode. Skills are\n written to compose, not just to exist.\n- **A generated knowledge graph with drift gates** — the catalog (nodes, ownership,\n ordered scenario edges) is exported from the markdown and CI-checked against it,\n so prose and structure provably cannot diverge. A pre-tool hook blocks\n confabulated skill names at invocation time.\n- **Adversarial review** — scenario content is periodically challenged by parallel\n expert-panel agent runs (consistency, currency, decision value), and fixes are\n verified by retrieval smoke tests before merging.\n- **Static security screening** — third-party skills are scanned for prompt\n injection, hidden Unicode, and exfiltration sinks before they enter the library\n (taxonomy adapted from NVIDIA/SkillSpector).\n- **Freshness discipline** — volatile facts (framework versions, model pricing)\n live in per-skill `data/` files refreshed by script, and prose is gated against\n its own data to stop version rot.\n\n## Standards & Compatibility\n\n- **Specification**: skills follow the [Agent Skills specification](https://agentskills.io/specification) —\n portable directories, YAML frontmatter, progressive disclosure.\n- **Platforms**: Claude Code (`~/.claude/skills`) and Codex CLI (`~/.agents/skills`); the same\n directories work in both.\n- **Self-containment**: every skill works detached from this repo — own `data/`, own `scripts/`,\n symlink-safe path resolution — so a single copied folder is a working install.\n\n## Scope\n\nThis is a curated public subset of a larger private library. Project-specific, client-scoped,\nmarketing, startup, and legal skills are not published, and skills depending on internal\nlibraries are excluded because they cannot be used outside their origin repo.\n\n## Contributing\n\nIssues and PRs welcome — see [CONTRIBUTING.md](CONTRIBUTING.md). Useful contributions:\ncorrecting something that has gone stale, adding a reference that saved you time, or\nreporting a skill that misfires.\n\n## License\n\nMIT — see [LICENSE](LICENSE). Use freely, including commercially.\n",
"bytes": 6896,
"sha": "f0e5b4a33b5d914bb9ceb29ab3d86ecbebef5f6de9bab9abd60c7c992805c603",
"repo_slug": "vasilyu1983/ai-agents-public",
"fonte": "repo",
"truncated": false,
"api": "https://agentalog.com/api/listings/skl_vasilyu1983_ai_agents_public_software_ar_8656d2e1/readme"
}