{"id":51298818,"url":"https://github.com/jomeswang/agnes-ai-skill","last_synced_at":"2026-06-30T17:02:19.061Z","repository":{"id":361837783,"uuid":"1256061018","full_name":"jomeswang/agnes-ai-skill","owner":"jomeswang","description":"Agnes AI skill for text, image, and video APIs with persistent auth and OpenAI-style workflows","archived":false,"fork":false,"pushed_at":"2026-06-11T12:11:42.000Z","size":25670,"stargazers_count":32,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2026-06-11T13:26:14.137Z","etag":null,"topics":["agent-skills","agnes-ai","ai-agent-skills","api-integration","claude-code-skills","codex-skills","image-generation","multimodal-ai","openclaw-skills","video-generation"],"latest_commit_sha":null,"homepage":"https://github.com/jomeswang/agnes-ai-skill","language":"Swift","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/jomeswang.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-06-01T12:28:55.000Z","updated_at":"2026-06-11T12:11:50.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/jomeswang/agnes-ai-skill","commit_stats":null,"previous_names":["jomeswang/agnes-ai-skill"],"tags_count":1,"template":false,"template_full_name":null,"purl":"pkg:github/jomeswang/agnes-ai-skill","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/jomeswang%2Fagnes-ai-skill","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/jomeswang%2Fagnes-ai-skill/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/jomeswang%2Fagnes-ai-skill/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/jomeswang%2Fagnes-ai-skill/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/jomeswang","download_url":"https://codeload.github.com/jomeswang/agnes-ai-skill/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/jomeswang%2Fagnes-ai-skill/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":34975672,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-06-30T02:00:05.919Z","response_time":92,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["agent-skills","agnes-ai","ai-agent-skills","api-integration","claude-code-skills","codex-skills","image-generation","multimodal-ai","openclaw-skills","video-generation"],"created_at":"2026-06-30T17:02:18.363Z","updated_at":"2026-06-30T17:02:19.055Z","avatar_url":"https://github.com/jomeswang.png","language":"Swift","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Agnes AI Free Text, Image \u0026 Video Generation Skill\n\n[![MIT License](https://img.shields.io/badge/license-MIT-green.svg)](./LICENSE)\n[![Agent Skill](https://img.shields.io/badge/Agent%20Skill-SKILL.md-blue)](./SKILL.md)\n[![Models](https://img.shields.io/badge/models-text%20%7C%20image%20%7C%20video-black)](https://agnes-ai.com/doc)\n[![Agnes AI](https://img.shields.io/badge/platform-Agnes%20AI-ff6b3d)](https://platform.agnes-ai.com/)\n[![Chinese README](https://img.shields.io/badge/docs-%E4%B8%AD%E6%96%87-blue)](./README.zh-CN.md)\n\nAgnes AI Skill is a free multimodal Agent Skill for Codex, Claude Code,\nOpenClaw, Claude Desktop, Hermes, WorkBuddy, Cherry Studio, Opencode, and\nsimilar AI agent tools. It helps agents use Agnes 2.0 APIs for text generation,\nimage generation, image editing, text-to-video, image-to-video, and API\nintegration through a single root `SKILL.md`.\n\nUse this Agnes skill when you want an Agnes AI skill, image generation skill,\nvideo generation skill, text to image workflow, image to video workflow, free AI API\nworkflow, or free AI model workflow that fits Codex skills, Claude Code skills,\nand OpenClaw skills.\n\n中文：Agnes AI 免费文本 / 生图 / 图片编辑 / 生视频 Agent Skill，支持 Codex、\nClaude Code、OpenClaw、Claude Desktop、Hermes、WorkBuddy、Cherry Studio、\nOpencode 等工具。\n\n## Quick Start\n\nInstall the skill with a repository-aware skills installer:\n\n```bash\nnpx skills add jomeswang/agnes-ai-skill -g\n```\n\nOr copy this sentence to your agent:\n\n```text\nPlease read and install Agnes AI Skill: https://github.com/jomeswang/agnes-ai-skill\n```\n\nAfter installation, create an Agnes API key at\n[platform.agnes-ai.com/settings/apiKeys](https://platform.agnes-ai.com/settings/apiKeys)\nand configure it as `AGNES_API_KEY`. Then tell the agent `generate an image`,\n`generate a video`, `edit this image`, or `integrate Agnes API`, and it can\nchoose the proper Agnes model and execution path automatically.\n\nThis skill helps agents:\n\n- set up Agnes AI and persist an `AGNES_API_KEY`\n- use `agnes-2.0-flash` for chat, coding, streaming, and tool calling\n- use `agnes-image-2.0-flash` and `agnes-image-2.1-flash` for AI image\n  generation, image editing, and image-to-image workflows\n- use `agnes-video-v2.0` for text-to-video, image-to-video, and async video\n  generation\n- prefer the companion `agnes-ai-cli` for live requests\n\nCompatible tools: Codex, Claude Code, Claude Desktop, OpenClaw, Hermes,\nWorkBuddy, Cherry Studio, Opencode, Kimi Work, and other Agent Skills /\nSKILL.md-compatible environments.\n\n![Agnes AI frontier models hero](./assets/images/agnes-frontier-models-banner.jpg)\n\n## Important Announcement\n\n**Agnes 2.0 full-modal model APIs are officially open for free global calls.**\n\n- No fixed end date, full-modal access, and free API calls within RPM 20\n- Register on the official platform, create a key, and call directly\n- Text, image, and video are all supported\n- Models keep upgrading while free access is maintained\n\n**Official platform:** https://platform.agnes-ai.com\n\n\u003e The live official model docs also include pricing sections. For cost, quota,\n\u003e or commercial use, check the current official docs first.\n\nThis repository packages a single root `SKILL.md` so coding agents can quickly:\n\n- get and persist an Agnes API key\n- use `agnes-2.0-flash` for chat, coding, streaming, and tool calling\n- use `agnes-image-2.0-flash` and `agnes-image-2.1-flash` for image generation\n  and editing\n- use `agnes-video-v2.0` for asynchronous video generation and polling\n\nIt is designed for the exact pitch that makes Agnes easy to try:\n\n- one provider for text, image, and video\n- public free-access positioning that lowers experimentation friction\n- agent, creative, and prototyping workflows where repeated calls matter\n\nThe skill stays intentionally lightweight. It teaches agents how to make Agnes\nAPI calls successfully without copying the full docs into the repository.\n\n## Install\n\nWith a repository-aware skills installer:\n\n```bash\nnpx skills add jomeswang/agnes-ai-skill -g\n```\n\nBecause this repository uses a single root `SKILL.md`, installers that support\nrepository-root skills can discover it directly.\n\n### Verified Install Path\n\nThis repository has been validated with:\n\n```bash\nnpx skills add jomeswang/agnes-ai-skill --list\nnpx skills add jomeswang/agnes-ai-skill --agent codex --yes\n```\n\nThe repository is discoverable as a single root-level skill named\n`agnes-ai-skill`.\n\n## Companion CLI\n\nThe published companion execution layer is:\n\n- npm: [`agnes-ai-cli`](https://www.npmjs.com/package/agnes-ai-cli)\n- GitHub: [`jomeswang/agnes-ai-cli`](https://github.com/jomeswang/agnes-ai-cli)\n\nInstall it when you want stable commands instead of hand-written requests:\n\n```bash\nnpm install -g agnes-ai-cli\nagnes --help\n```\n\n## Model Guide\n\nUse the repository skill with these defaults:\n\n- `agnes-2.0-flash`\n  - chat, coding, streaming, tool calling, and agent workflows\n  - default when `text chat` runs without `--model`\n- `agnes-image-2.1-flash`\n  - default for new text-to-image and image-to-image work\n  - strongest fit for denser layouts, richer detail, and better semantic\n    alignment\n  - default when `image text2img`, `image img2img`, or `image compose` runs\n    without `--model`\n- `agnes-image-2.0-flash`\n  - better when you explicitly need its documented `tags: [\"img2img\"]` flow,\n    multi-image composition, or `seed`-based reproducibility\n- `agnes-video-v2.0`\n  - text-to-video, image-to-video, multi-image guided video, keyframes, and\n    asynchronous polling\n  - current default when any `video` generate command runs without `--model`\n\nThe official docs also give each model fairly different best practices:\n\n- Image 2.1 leans into high-information-density visuals and composition\n  preservation\n- Image 2.0 is more explicit about edit/composition workflows, response fields,\n  and OpenAI Images-style compatibility\n- Video 2.0 is task-based and documents multiple generation modes, task states,\n  result polling, and frame-count constraints\n\n## Showcase\n\nThe strongest outside prompt libraries all do the same three things well:\nlead with a preview, keep the prompt compact enough to scan, and organize\nexamples by outcome instead of by raw API parameter lists. This gallery follows\nthat pattern with 9 compact cases regenerated with Agnes on June 1, 2026. The\nimage prompts were tightened toward the shorter, more editorial style seen in\n`awesome-gpt-image-2`, the video prompts were rebuilt around ad-film,\nanimation, and cinematic beats inspired by `awesome-seedance`, and the HTML\ncases were regenerated as brighter, more public-facing product experiences.\n\n### Image Cases\n\n`agnes-image-2.1-flash`\n\n| Preview | Case | Model | Prompt recipe |\n| --- | --- | --- | --- |\n| [![Y2K golden hour portrait](./assets/images/y2k-golden-hour.png)](./assets/images/y2k-golden-hour.png) | Y2K golden hour portrait | `agnes-image-2.1-flash` | Candid blonde portrait, baby-pink velour tracksuit, butterfly clips, glossy lips, palm trees, warm golden hour, real skin texture, soft film grain. |\n| [![Pencil editorial fashion](./assets/images/pencil-editorial-fashion.png)](./assets/images/pencil-editorial-fashion.png) | Pencil editorial fashion | `agnes-image-2.1-flash` | Minimal monochrome pencil-sketch fashion portrait, round sunglasses, rolled white shirt, denim overalls, combat boots, burnt-orange circle, indie magazine composition. |\n| [![Brand envelope perfume ad](./assets/images/brand-envelope-perfume.png)](./assets/images/brand-envelope-perfume.png) | Brand-envelope perfume ad | `agnes-image-2.1-flash` | Dusty-rose brand world, travertine pedestal, translucent perfume bottle, matte cream paper curves, warm studio light, quiet-luxury beauty campaign. |\n\n### Video Cases\n\n`agnes-video-v2.0`\n\n| Preview | Case | Model | Prompt recipe |\n| --- | --- | --- | --- |\n| [![Perfume ad film first frame](./assets/video-previews/ad-film-10s.jpg)](./assets/videos/ad-film-10s.mp4) | Perfume ad film | `agnes-video-v2.0` | Ten-second luxury commercial with macro bottle details, orbiting reflections, atomizer tension beat, champagne-gold glow, and a final hero reveal built for premium beauty launch pages. |\n| [![Mechanical otter animation short first frame](./assets/video-previews/animation-film-10s.jpg)](./assets/videos/animation-film-10s.mp4) | Mechanical otter animation short | `agnes-video-v2.0` | Ten-second animated short: fearless otter pilot bursts through a clockwork engine room, expressive face, hand-painted adventure energy, cinematic action timing, family-film clarity. |\n| [![Desert cinematic film scene first frame](./assets/video-previews/cinema-film-10s.jpg)](./assets/videos/cinema-film-10s.mp4) | Desert cinematic film scene | `agnes-video-v2.0` | Ten-second widescreen film moment: lone traveler crosses a dust-heavy desert test site, sculptural machine in the background, restrained grading, prestige sci-fi drama mood. |\n\n### App Cases\n\n`agnes-2.0-flash`\n\nSingle-file HTML demos generated from Agnes text prompts and saved in\n[`examples/apps`](./examples/apps).\n\n| Preview | Case | Model | Prompt recipe |\n| --- | --- | --- | --- |\n| [![Agnes AI official-style homepage](./assets/apps/cinematic-ai-landing.jpg)](./examples/apps/cinematic-ai-landing.html) | Agnes AI official-style homepage | `agnes-2.0-flash` | Fresh one-file official product homepage for Agnes AI, bright premium palette, strong platform nav, enterprise hero copy, high-clarity call-to-action hierarchy. |\n| [![Lantern Sprint mini-game](./assets/apps/beijing-map-ui.jpg)](./examples/apps/beijing-map-ui.html) | Lantern Sprint mini-game | `agnes-2.0-flash` | Fresh browser mini-game with bright festival palette, instant start state, arcade score loop, keyboard and touch controls, polished one-file mobile-friendly presentation. |\n| [![Golden Hour Stories promo page](./assets/apps/watch-experience.jpg)](./examples/apps/watch-experience.html) | Golden Hour Stories promo page | `agnes-2.0-flash` | Fresh campaign landing page for Agnes image and video outputs, soft daylight tones, editorial serif hero, luxury launch framing, clear promotional storytelling. |\n\n## Why Agnes\n\nAgnes is most interesting when one workflow needs all three layers together:\n\n- text for planning, coding, prompting, and agent loops\n- image for marketing, e-commerce, and creative visual generation\n- video for storyboards, product demos, motion tests, and short-form content\n\nThe supplied public writeups consistently frame Agnes as useful for:\n\n- rapid AI product prototyping\n- high-frequency agent workflows where repeated model calls matter\n- frontend or HTML generation\n- marketing and e-commerce creatives\n- ad, storyboard, and cinematic short-video iteration\n\nThis skill turns that platform surface into one reusable installation target for\nCodex and other SKILL.md-compatible agents, with guidance that helps the agent\nchoose the right Agnes model and authenticate cleanly.\n\n## What It Does\n\n- Platform and auth flow from the Agnes quickstart docs\n- API key creation via the Agnes platform settings page\n- Persistent `AGNES_API_KEY` setup for future sessions\n- OpenAI-style request patterns for text and image endpoints\n- Asynchronous task workflow for video generation\n- Practical use cases reinforced by the supplied public writeups\n\n## Why The Free Angle Matters\n\n- Agnes is unusually compelling when one platform covers text, image, and video\n  together.\n- The strongest growth hook in its public messaging is not just quality, but\n  the promise of lower-cost or broadly free experimentation.\n- That matters most for agents, prototypes, content pipelines, and repeated\n  A/B-style creative iteration.\n- In practice, treat this as a major adoption advantage, while still verifying\n  the current live billing terms before promising zero cost.\n\n## Model Quick Reference\n\n| Model | Best for | Endpoint | Required fields | Special fields / caveats |\n| --- | --- | --- | --- | --- |\n| `agnes-2.0-flash` | chat, coding, streaming, tool calling, agent loops | `/v1/chat/completions` | `model`, `messages` | OpenAI-style `tools`, `tool_choice`, `stream` |\n| `agnes-image-2.1-flash` | new text-to-image, image-to-image, denser layouts, composition-preserving edits | `/v1/images/generations` | `model`, `prompt` | `size`, `extra_body.image`, `extra_body.response_format`; strongest fit for high-information-density scenes |\n| `agnes-image-2.0-flash` | edit-heavy workflows, multi-image composition, compatibility-style image flows | `/v1/images/generations` | `model`, `prompt` | often pair with `tags: [\"img2img\"]`; supports `seed`, `extra_body.image`, `extra_body.response_format` |\n| `agnes-video-v2.0` | text-to-video, image-to-video, multi-image guided video, keyframes | create: `/v1/videos`; poll: `/agnesapi?video_id={video_id}` | `model`, `prompt` | asynchronous task workflow; create returns `taskId` and `videoId`; use `videoId` for polling, while older `taskId` polling remains legacy-compatible |\n\n## Pricing / Operational Caveats\n\n- The official Image 2.0, Image 2.1, and Video 2.0 docs currently include\n  pricing sections.\n- Some public writeups still frame Agnes as broadly free or indefinitely free.\n- The Video 2.0 docs are also operationally time-sensitive: the same page can\n  mix concrete price figures with \"pricing to be announced\" style notes.\n- Treat all pricing, free-tier, and billing claims as live-doc verified only.\n\n## Safety Model\n\n- The skill checks for `AGNES_API_KEY` before live requests\n- If the key is missing, it points the user to the official Agnes quickstart\n  and API key page instead of guessing\n- If the user provides a key and wants it remembered, the skill persists\n  `AGNES_API_KEY` in the correct shell rc file for future sessions\n- Live payloads and response handling stay grounded in Agnes docs and real API\n  behavior, not only in marketing copy\n\n## Manual Install\n\nCopy this repository into any standard skills location supported by your agent,\nfor example:\n\n- Codex: `~/.codex/skills/agnes-ai-skill`\n- Claude Code: `~/.claude/skills/agnes-ai-skill`\n- Cursor: `~/.cursor/skills/agnes-ai-skill`\n\n## What Agents Learn\n\n- How to detect missing Agnes auth before making live calls\n- How to guide the user to create an API key\n- How to persist `AGNES_API_KEY` in shell startup files for future sessions\n- How to choose between Agnes text, image, and video models\n- How to make the smallest reliable live request first\n- How to poll Agnes video tasks until they complete\n\n## Discovery Notes\n\n- GitHub repository: [jomeswang/agnes-ai-skill](https://github.com/jomeswang/agnes-ai-skill)\n- Public repository topics:\n  `agent-skills`, `ai-agent-skills`, `codex-skills`, `multimodal-ai`, `agnes-ai`\n- These topics improve discoverability across GitHub-linked skill directories\n  and crawler-based ecosystems.\n- The repository is ready for third-party skill hub submission, including\n  ClawHub-style marketplaces that read `SKILL.md` metadata.\n\n## Primary Sources\n\n- Agnes quickstart: [https://agnes-ai.com/doc/quickstart](https://agnes-ai.com/doc/quickstart)\n- API key settings:\n  [https://platform.agnes-ai.com/settings/apiKeys](https://platform.agnes-ai.com/settings/apiKeys)\n- Text model:\n  [https://agnes-ai.com/doc/agnes-20-flash](https://agnes-ai.com/doc/agnes-20-flash)\n- Image 2.0:\n  [https://agnes-ai.com/doc/agnes-image-20-flash](https://agnes-ai.com/doc/agnes-image-20-flash)\n- Image 2.1:\n  [https://agnes-ai.com/doc/agnes-image-21-flash](https://agnes-ai.com/doc/agnes-image-21-flash)\n- Video:\n  [https://agnes-ai.com/doc/agnes-video-v20](https://agnes-ai.com/doc/agnes-video-v20)\n\n## Notes\n\n- The public materials supplied with this repository describe Agnes as offering\n  free access to its core multimodal APIs as of June 1, 2026. Treat pricing and\n  promotion details as time-sensitive and verify them in the platform if cost\n  matters.\n- ClawHub publishing requires a separate ClawHub login or publish token plus a\n  GitHub OAuth grant.\n- This repository is released under the MIT License.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjomeswang%2Fagnes-ai-skill","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fjomeswang%2Fagnes-ai-skill","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjomeswang%2Fagnes-ai-skill/lists"}