$git clone https://github.com/theagenticguy/agentic-pluginsPublic Claude Code plugin marketplace — installable skills and agents for AI coding agents. Drop-in capabilities you add with one command.
| 1 | # agentic-plugins |
| 2 | |
| 3 | A public [Claude Code](https://code.claude.com) plugin marketplace — installable skills and agents |
| 4 | for AI coding agents. |
| 5 | |
| 6 | ## Install |
| 7 | |
| 8 | Add the marketplace, then install a plugin: |
| 9 | |
| 10 | ```bash |
| 11 | /plugin marketplace add theagenticguy/agentic-plugins |
| 12 | /plugin install agentic-skills@agentic-plugins |
| 13 | /plugin install plugin-gardener@agentic-plugins |
| 14 | ``` |
| 15 | |
| 16 | ## Plugins |
| 17 | |
| 18 | ### agentic-skills |
| 19 | |
| 20 | Flagship skills for AI coding agents. |
| 21 | |
| 22 | | Skill | What it does | |
| 23 | | ------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | |
| 24 | | `workbench-builder` | Builds disposable localhost workbenches — Flask + `sqlite3` + htmx + Server-Sent Events apps that give an agentic coding/eval/PR/data session a live UI with no build step, no `npm install`, and no deploy. The signature move is a **two-way human↔agent loop** over one SQLite file: the human acts in the browser (htmx `POST` → SQLite → fragment + SSE invalidation), the agent acts from the terminal (`httpx` → SQLite → SSE), both seeing the same state update live with no reload. Ships recipes for eval viewers, PR review rooms, document-redline surfaces, trace replays, and refactor cockpits. | |
| 25 | |
| 26 | **Bundled MCP servers** (`plugins/agentic-skills/.mcp.json`) — research and documentation tools that |
| 27 | back the skills. Each reads its key from an environment variable; the marketplace entry is |
| 28 | `strict: false`, so a missing key never blocks install — the server is simply skipped. |
| 29 | |
| 30 | | Server | Purpose | Env var | |
| 31 | | ------------ | --------------------------- | ------------------ | |
| 32 | | context7 | Up-to-date library/API docs | `CONTEXT7_API_KEY` | |
| 33 | | deepwiki | GitHub repo Q&A | *(none)* | |
| 34 | | brave-search | Web search | `BRAVE_API_KEY` | |
| 35 | | tavily | Web search + extraction | `TAVILY_API_KEY` | |
| 36 | | exa | Neural web search | `EXA_API_KEY` | |
| 37 | | you | You.com search + research | `YDC_API_KEY` | |
| 38 | | awsknowledge | AWS documentation | *(none)* | |
| 39 | |
| 40 | ### plugin-gardener |
| 41 | |
| 42 | Catalog hygiene for Claude Code skill/agent plugins. Runs audits — inventory, per-skill scoring, |
| 43 | embedding-based collision detection, and HDBSCAN taxonomy checks — and proposes PR-style |
| 44 | resolutions without auto-applying. |
| 45 | |
| 46 | ## Repository layout |
| 47 | |
| 48 | ```text |
| 49 | .claude-plugin/marketplace.json # marketplace registry — lists every plugin |
| 50 | plugins/ |
| 51 | agentic-skills/ # flagship plugin |
| 52 | .claude-plugin/plugin.json |
| 53 | .mcp.json # bundled MCP servers |
| 54 | skills/ |
| 55 | plugin-gardener/ # catalog-hygiene plugin |
| 56 | .claude-plugin/plugin.json |
| 57 | skills/ references/ scripts/ |
| 58 | tools/ # repo-wide validators and scaffolding |
| 59 | ``` |
| 60 | |
| 61 | ## Development |
| 62 | |
| 63 | Requires [mise](https://mise.jdx.dev/) and [uv](https://docs.astral.sh/uv/). |
| 64 | |
| 65 | ```bash |
| 66 | mise install # node, markdownlint-cli2, dprint |
| 67 | mise run fmt # format all markdown and JSON |
| 68 | mise run fmt:check # check formatting (CI-friendly) |
| 69 | mise run lint # lint all markdown |
| 70 | mise run lint:md:fix # lint with auto-fix |
| 71 | mise run validate # frontmatter + size + refs + freshness + versions |
| 72 | mise run build # full gate: lint + fmt:check + validate |
| 73 | mise run bump -- <plugin> patch # bump one plugin's version |
| 74 | mise run init:skill -- <plugin> <name> 'Description. Use when...' # scaffold a skill |
| 75 | ``` |
| 76 | |
| 77 | CI runs `mise run build` on every push and pull request. |
| 78 | |
| 79 | ## License |
| 80 | |
| 81 | [MIT](LICENSE) |