A curated repository of reusable agent skills, references, and helper scripts.
This repo is designed to store portable skill folders that can be copied into a compatible agent runtime, or adapted for other agent systems such as Codex, Claude-oriented wrappers, OpenClaw, and similar tool-using assistants.
skills_for_agent/
├── README.md
├── skills/
│ └── <skill-name>/
│ ├── SKILL.md
│ ├── agents/
│ ├── scripts/
│ ├── references/
│ └── assets/
└── *.md
skills/: primary home for reusable skill foldersSKILL.md: core instructions and trigger metadata for a skillagents/: optional agent-specific metadata such asopenai.yamlscripts/: helper utilities the skill can run directlyreferences/: deeper docs loaded only when needed- root-level
*.md: standalone guides or legacy references that are not yet packaged as full skills
Copy a skill folder into ~/.codex/skills/:
mkdir -p ~/.codex/skills
cp -R skills/imgui-repo-navigator ~/.codex/skills/For other agent frameworks, keep the skill folder intact and install it into that agent's skill/plugin directory, or point the agent at the folder directly if supported.
Minimum portable unit:
- the entire skill folder
SKILL.md- any referenced
scripts/andreferences/
Each skill may declare its own runtime requirements. For example:
skills/imgui-repo-navigatorrequirespython3- upstream GitHub fallback also requires network access
| Skill | Purpose | Key Dependencies | Notes |
|---|---|---|---|
github-stars-organizer |
Audit, classify, clean up, and reorganize GitHub stars into reusable themed GitHub Lists | python3, GitHub PAT, network |
Supports export, cleanup proposals, bulk unstar, list sync, list reuse for the 32-list cap, and old-list cleanup |
gemini-web-image |
Generate or vary a single image through Gemini Web with an isolated local Chrome login | bun or npx bun, Chrome/Chromium, interactive login |
Unofficial reverse-engineered web flow, image-first, text mode only for diagnostics |
imgui-repo-navigator |
Find Dear ImGui APIs, examples, backend files, and docs from a local checkout or upstream GitHub | python3, optional network |
Local-first, GitHub fallback |
n8n-api-v1 |
Operate n8n API v1 with authenticated request patterns and endpoint references | n8n API access, optional curl/jq |
Defaults to local localhost:5678/api/v1, base URL overridable |
open-source-radar |
Find current hot repos and open-source trend signals for any domain, niche, or keyword | live GitHub access, optional ecosystem sources | General-purpose radar workflow with GitHub-first source priority and reproducible query trails |
pi-xhs-card |
Build Xiaohongshu card decks in a local web studio with JSON-backed editing, template switching, and PNG export | node, curl, playwright, Chrome or Chromium |
Browser-based workflow, overflow-safe exports, starter studio included |
project-continuity-manager |
Keep large projects coherent across many sessions and agents with a durable docs/agent/ memory layer and validated handoffs |
python3, optional git |
Initializes templates, checks docs health, creates handoffs, validates handoffs, and checks handoff freshness |
Recommended rules for new skills added to this repo:
- Keep each skill self-contained in its own folder under
skills/ - Put trigger logic and operating instructions in
SKILL.md - Put long reference material in
references/ - Put executable helpers in
scripts/ - Keep agent-specific metadata optional and non-blocking
- Prefer cross-platform scripts when possible
- Prefer local/project-aware lookup first, then upstream/network fallback when that improves portability
Use the ImGui skill after installation:
Use $imgui-repo-navigator to find the correct Win32 + DX9 backend integration flow.
Use the GitHub stars organizer after installation:
Use $github-stars-organizer to export my stars, propose theme-based groupings, and sync the final plan into GitHub Lists.
Use the open-source radar skill after installation:
Use $open-source-radar to find this month's hot and worth-following open-source projects in AI agents.
Use the PI XHS Card skill after installation:
Use $pi-xhs-card to generate a Xiaohongshu card deck, let me edit it in a browser, and export a full PNG set without cropped content.
Use the Gemini Web image skill after installation:
Use $gemini-web-image to generate one image through Gemini Web from a prompt or reference image.
Or run its helper directly:
python skills/imgui-repo-navigator/scripts/find_imgui_targets.py win32 dx9
python skills/imgui-repo-navigator/scripts/find_imgui_targets.py --mode upstream --ref master docking