Back to search
omer-plugins
MarketplaceWorks with Claude CodeGitHubIncomplete
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Checked today
Badge for your README
Markdown
[](/listings/omer-yehuda--plugins)HTML
<a href="/listings/omer-yehuda--plugins"><img src="/api/listings/omer-yehuda--plugins/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
Run this in Claude Code (v2.1.275 or later):
/plugin install army --marketplace omer-yehuda/plugins /plugin install toon-shorts --marketplace omer-yehuda/plugins /plugin install court --marketplace omer-yehuda/plugins /plugin install customs --marketplace omer-yehuda/plugins /plugin install er --marketplace omer-yehuda/plugins /plugin install atc --marketplace omer-yehuda/plugins /plugin install studio --marketplace omer-yehuda/plugins /plugin install critics --marketplace omer-yehuda/plugins /plugin install skill-installer --marketplace omer-yehuda/plugins /plugin install mission-control --marketplace omer-yehuda/plugins /plugin install academy --marketplace omer-yehuda/plugins /plugin install archive --marketplace omer-yehuda/plugins /plugin install newsroom --marketplace omer-yehuda/plugins /plugin install premiere --marketplace omer-yehuda/plugins
Install only the ones you want.
<details> <summary>Older Claude Code versions</summary>
/plugin marketplace add omer-yehuda/plugins /plugin install army@omer-plugins /plugin install toon-shorts@omer-plugins /plugin install court@omer-plugins # same pattern for customs, er, atc, studio, critics, skill-installer, mission-control, academy, archive, newsroom, premiere
</details>
Restart Claude Code (or run `/reload-plugins`), then type:
raise the army build a landing page for my habit-tracking app make me an animated short about the man who sold the Eiffel Tower put it on trial: should we move from REST to GraphQL? /customs:inspect https://github.com/<owner>/<repo> admit this bug: checkout total is off by one cent /atc:claim src/billing/** (invoice refactor) studio, make a video about what a Claude skill is let the critics taste it install the pdf skill from anthropics/skills mission control, are we go? send this skill to the academy ask the archive why we chose REST newsroom, write the edition premiere, make a film about the last lighthouse keeper
To update later, run `/plugin marketplace update omer-plugins`.
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
.claude-plugin/marketplace.json · Manifest · 213 lines
- {
- "name": "omer-plugins",
- "owner": {
- "name": "Omer Yehuda",
- "url": "https://github.com/omer-yehuda"
- },
- "metadata": {
- "description": "A curated marketplace of tested Claude Code plugins by Omer Yehuda: Army (multi-agent orchestration), Toon Shorts (animated shorts), Code Court (decisions on trial), Customs (a first-filter scan of skills before you install), Bug ER (minimal-incision bug fixes), ATC (air traffic control for parallel sessions), Studio (AI film crew for explainer videos), The Critics' Table (3-star review), Skill Installer (install skills from GitHub), Mission Control (go or no-go before a release), The Academy (exams for skills), The Archive (one memory for every plugin), The Newsroom (release notes from commits) and Premiere (realistic short films)."
- },
- "plugins": [
- {
- "name": "army",
- "source": "./plugins/army",
- "description": "A chief of staff that commands 18 specialist units (intel, engineering, defense, legal, comms, data, sales and more) as subagents, with 110+ on-demand specialist playbooks, cost-aware model routing, safety gates and local lessons-learned memory.",
- "category": "productivity",
- "keywords": [
- "orchestration",
- "subagents",
- "multi-agent",
- "security",
- "code-review",
- "research",
- "legal"
- ]
- },
- {
- "name": "toon-shorts",
- "source": "./plugins/toon-shorts",
- "description": "Animated short creator: turns a true story into an original voiced, captioned 1080x1920 cartoon MP4 for Shorts/TikTok/Reels, with toon characters, parallax sets, karaoke captions (Hebrew RTL), and a self-reviewed preview.",
- "category": "creative",
- "keywords": [
- "video",
- "shorts",
- "animation",
- "tiktok",
- "reels",
- "storytelling",
- "tts",
- "hebrew"
- ]
- },
- {
- "name": "court",
- "source": "./plugins/court",
- "description": "Put any technical decision on trial. Opposing counsel argue each option from shared evidence, a judge who never saw your opinion rules, and the verdict names the conditions that would overturn it.",
- "category": "productivity",
- "keywords": [
- "decisions",
- "architecture",
- "adr",
- "subagents",
- "debate",
- "trade-offs"
- ]
- },
- {
- "name": "customs",
- "source": "./plugins/customs",
- "description": "Border control for skills and plugins: every skill needs a passport. Scans a plugin or skill (local path, GitHub URL or installed plugin) before you trust it and stamps it CLEARED, SECONDARY INSPECTION or DENIED. A first filter, not a guarantee.",
- "category": "security",
- "keywords": [
- "security",
- "supply-chain",
- "skills",
- "plugins",
- "audit",
- "scanner",
- "prompt-injection"
- ]
- },
- {
- "name": "er",
- "source": "./plugins/er",
- "description": "Your bug gets admitted, triaged, diagnosed and operated on with the smallest possible incision. A triage nurse, an attending, a surgeon and a discharge check, with a hard incision budget so a one-line bug never becomes a 40-file rewrite.",
- "category": "development",
- "keywords": [
- "debugging",
- "bugfix",
- "minimal-diff",
- "regression-test",
- "triage",
- "subagents"
- ]
- },
- {
- "name": "atc",
- "source": "./plugins/atc",
- "description": "Air traffic control for parallel Claude Code sessions: claim lanes, avoid collisions. Sessions claim lanes (file globs); a hook asks before one session edits a file in another's lane, and a clearance skill splits a big task across non-overlapping parallel pilots. Coordination, not a sandbox.",
- "category": "productivity",
- "keywords": [
- "parallel",
- "subagents",
- "worktrees",
- "coordination",
- "file-locking",
- "multi-session"
- ]
- },
- {
- "name": "studio",
- "source": "./plugins/studio",
- "description": "A film crew of AI agents that turns a topic into a finished explainer video, vertical (1080x1920) or wide (1920x1080): writer, director, test-screening critic and editor produce a kinetic-typography MP4 with captions, optional free local AI narration and upload metadata. The trend desk films a trending tool from real screenshots and makes a wide video plus a Short. Local rendering with Pillow and ffmpeg.",
- "category": "creative",
- "keywords": [
- "video",
- "shorts",
- "explainer",
- "motion-graphics",
- "kinetic-typography",
- "agents",
- "captions"
- ]
- },
- {
- "name": "critics",
- "source": "./plugins/critics",
- "description": "Three independent critics (presentation, kitchen, recipe) taste your work before it ships, then a head critic awards up to 3 stars with a short fix-before-serving list. Re-taste after fixes to see stars go up.",
- "category": "development",
- "keywords": [
- "code-review",
- "review",
- "quality",
- "ux",
- "performance",
- "subagents"
- ]
- },
- {
- "name": "skill-installer",
- "source": "./plugins/skill-installer",
- "description": "Install Claude skills straight from GitHub: browse a skills repo or a plugin marketplace, dry-run first, get a quick risk check, then install into ~/.claude/skills or the project with an atomic copy and a backup on update.",
- "category": "productivity",
- "keywords": [
- "skills",
- "installer",
- "github",
- "marketplace",
- "package-manager"
- ]
- },
- {
- "name": "mission-control",
- "source": "./plugins/mission-control",
- "description": "Go or no-go before every release. A script collects the release facts (versions, changelogs, dirty tree), four stations report, and a flight director who never saw your opinion rules GO, HOLD or NO-GO. It never tags, pushes or deploys.",
- "category": "productivity",
- "keywords": [
- "release",
- "go-no-go",
- "checklist",
- "versioning",
- "changelog",
- "subagents"
- ]
- },
- {
- "name": "academy",
- "source": "./plugins/academy",
- "description": "No skill ships without passing the exam. An examiner writes missions from a skill's promise, the script runs each one with and without the skill in headless Claude Code, and the report card says GRADUATED, RETAKE or EXPELLED with pass rate, tokens, cost and time.",
- "category": "development",
- "keywords": [
- "evals",
- "skills",
- "testing",
- "benchmark",
- "quality",
- "subagents"
- ]
- },
- {
- "name": "archive",
- "source": "./plugins/archive",
- "description": "One memory for every plugin. Searches the local records your other plugins already keep (court verdicts, bug charts, reviews, lessons, countdowns, exams) and answers with the file behind each claim. Read-only: it never moves or edits a record.",
- "category": "productivity",
- "keywords": [
- "memory",
- "search",
- "history",
- "knowledge",
- "citations",
- "subagents"
- ]
- },
- {
- "name": "newsroom",
- "source": "./plugins/newsroom",
- "description": "Every release deserves a headline. A reporter reads the commits, an editor writes release notes, a changelog draft, social posts and video metadata, and a fact-checker cuts every claim that has no commit behind it. Drafts only: it never posts.",
- "category": "productivity",
- "keywords": [
- "release-notes",
- "changelog",
- "announcements",
- "social",
- "writing",
- "subagents"
- ]
- },
- {
- "name": "premiere",
- "source": "./plugins/premiere",
- "description": "A film crew of AI agents for realistic short films. A director writes the shot list, a cinematographer writes one prompt per shot with a consistent look, you generate the stills with the image model you choose, a continuity check rejects bad frames, and ffmpeg cuts a graded, scored film with camera moves, in wide or vertical.",
- "category": "creative",
- "keywords": [
- "video",
- "film",
- "cinematic",
- "storyboard",
- "ffmpeg",
- "image-generation",
- "agents"
- ]
- }
- ]
- }
README.md · README · 269 lines
- <div align="center">
- <img src="docs/images/icon.png" width="84" alt="Army icon">
- # Omer Yehuda · Claude Code Plugins
- **A curated marketplace of tested plugins for [Claude Code](https://code.claude.com).**
- [](https://github.com/omer-yehuda/plugins/actions/workflows/validate.yml)
- [](LICENSE)
- 
- 
- 
- **[Army command center (live)](https://omer-yehuda.github.io/plugins/)** · [Toon Shorts](https://omer-yehuda.github.io/plugins/toon-shorts/) · [Order of battle](https://omer-yehuda.github.io/plugins/#units)
- </div>
- ---
- ## Quick start
- Run this in Claude Code (v2.1.275 or later):
- ```
- /plugin install army --marketplace omer-yehuda/plugins
- /plugin install toon-shorts --marketplace omer-yehuda/plugins
- /plugin install court --marketplace omer-yehuda/plugins
- /plugin install customs --marketplace omer-yehuda/plugins
- /plugin install er --marketplace omer-yehuda/plugins
- /plugin install atc --marketplace omer-yehuda/plugins
- /plugin install studio --marketplace omer-yehuda/plugins
- /plugin install critics --marketplace omer-yehuda/plugins
- /plugin install skill-installer --marketplace omer-yehuda/plugins
- /plugin install mission-control --marketplace omer-yehuda/plugins
- /plugin install academy --marketplace omer-yehuda/plugins
- /plugin install archive --marketplace omer-yehuda/plugins
- /plugin install newsroom --marketplace omer-yehuda/plugins
- /plugin install premiere --marketplace omer-yehuda/plugins
- ```
- Install only the ones you want.
- <details>
- <summary>Older Claude Code versions</summary>
- ```
- /plugin marketplace add omer-yehuda/plugins
- /plugin install army@omer-plugins
- /plugin install toon-shorts@omer-plugins
- /plugin install court@omer-plugins # same pattern for customs, er, atc, studio, critics, skill-installer, mission-control, academy, archive, newsroom, premiere
- ```
- </details>
- Restart Claude Code (or run `/reload-plugins`), then type:
- ```
- raise the army build a landing page for my habit-tracking app
- make me an animated short about the man who sold the Eiffel Tower
- put it on trial: should we move from REST to GraphQL?
- /customs:inspect https://github.com/<owner>/<repo>
- admit this bug: checkout total is off by one cent
- /atc:claim src/billing/** (invoice refactor)
- studio, make a video about what a Claude skill is
- let the critics taste it
- install the pdf skill from anthropics/skills
- mission control, are we go?
- send this skill to the academy
- ask the archive why we chose REST
- newsroom, write the edition
- premiere, make a film about the last lighthouse keeper
- ```
- To update later, run `/plugin marketplace update omer-plugins`.
- ## Plugin catalog
- | Plugin | What it does | Version |
- |---|---|---|
- | [**army**](plugins/army/README.md) | Multi-agent orchestration. A Chief of Staff commands 18 specialist units (intel, engineering, defense, legal, comms, data, sales and more) as subagents, backed by 110 on-demand field manuals. Session-bound safety gates, least-privilege tool kits, cost-aware model routing and local lessons-learned memory. | 1.2.0 |
- | [**toon-shorts**](plugins/toon-shorts/README.md) | Animated short creator. Turns a true story into an original voiced, captioned 1080x1920 cartoon MP4 for Shorts, TikTok and Reels: 10 ideas, retention script, free AI voice or your own, toon characters on parallax sets, karaoke captions with Hebrew RTL, and a preview Claude reviews before rendering. | 1.0.1 |
- | [**court**](plugins/court/README.md) | Code Court. Puts a technical decision on trial: opposing counsel argue each option from shared evidence, one rebuttal round, and a judge who never saw your opinion rules, with a dissent and the conditions that would overturn the verdict. | 1.0.0 |
- | [**customs**](plugins/customs/README.md) | Border control for skills and plugins. Scans a local folder, GitHub repo or installed plugin before you trust it (remote code, obfuscation, hidden Unicode, prompt injection, credential access, exfiltration, wildcard hooks) and stamps a passport: CLEARED, SECONDARY INSPECTION or DENIED. A first filter, not a guarantee. | 1.0.0 |
- | [**er**](plugins/er/README.md) | Bug ER. A triage nurse, attending, surgeon and discharge check fix a bug with a regression test first and the smallest possible incision, under a hard budget (default 3 files, 40 lines). | 1.0.0 |
- | [**atc**](plugins/atc/README.md) | ATC: air traffic control for parallel Claude sessions. Sessions claim lanes (file globs), a hook asks before one session edits another's lane, and `/atc:clear` splits a big task across non-overlapping parallel pilots in worktrees. | 1.0.0 |
- | [**studio**](plugins/studio/README.md) | An AI film crew. Writer, director, test-screening critic and editor turn a topic into a vertical or wide kinetic-typography explainer MP4 with captions, optional free local AI narration and upload metadata. The trend desk films a trending tool from real screenshots and makes a wide video plus a Short. Rendered locally with Pillow and ffmpeg. | 1.1.0 |
- | [**critics**](plugins/critics/README.md) | The Critics' Table. Three independent critics (presentation, kitchen, recipe) taste your work, a head critic awards up to 3 stars with a short fix-before-serving list, and a re-taste shows the stars go up. | 1.0.0 |
- | [**skill-installer**](plugins/skill-installer/README.md) | Skill Installer. Browse a skills repo or a plugin marketplace and install a skill from GitHub into `~/.claude/skills` or the project: dry run, quick risk check, atomic copy, backup on update, "did you mean" for typos. Adapted from OpenAI's Apache-2.0 Codex skill-installer. | 1.0.0 |
- | [**mission-control**](plugins/mission-control/README.md) | Mission Control. Go or no-go before a release: a script collects the facts (versions that disagree, changes without a changelog entry, a dirty tree), four stations report, and an independent flight director rules GO, HOLD or NO-GO. It never tags, pushes or deploys. | 1.0.0 |
- | [**academy**](plugins/academy/README.md) | The Academy. An examiner writes missions from a skill's promise, each one runs with and without the skill in headless Claude Code, and the script grades it GRADUATED, RETAKE or EXPELLED with pass rate, tokens, cost and time. | 1.0.0 |
- | [**archive**](plugins/archive/README.md) | The Archive. One read-only search over the records your other plugins keep (verdicts, bug charts, reviews, lessons, rulings, report cards). A librarian answers with the file behind every claim. | 1.0.0 |
- | [**newsroom**](plugins/newsroom/README.md) | The Newsroom. A reporter reads the commits, an editor drafts release notes, a changelog entry, social posts and video metadata, and a fact-checker cuts every claim with no commit behind it. Drafts only. | 1.0.0 |
- | [**premiere**](plugins/premiere/README.md) | Premiere. A film crew for realistic short films: director, cinematographer and continuity agents, stills from the image model you choose, and an ffmpeg cut with camera moves, one grade, a score and captions. | 1.0.0 |
- ---
- ## Featured: Army
- > Give Claude a mission, not a prompt.
- Say **"raise the army"** and Claude becomes a Chief of Staff. It asks its briefing questions once, writes an operation order, and dispatches specialist units in parallel, each on the most cost-effective model for its job. An independent Inspector General checks the result against your requirements before anything is reported back.
- [](https://omer-yehuda.github.io/plugins/)
- ```mermaid
- flowchart LR
- U([You]) -->|mission| HQ{{Chief of Staff}}
- HQ -->|briefing questions, once| U
- HQ --> R[Recon]
- HQ --> O[Ops & Planning]
- HQ --> I[Intel]
- HQ --> E[Engineering]
- HQ --> D[Defense]
- HQ --> L[Legal]
- HQ --> C[Comms]
- HQ --> X[Data · UX · Logistics · Treasury · Growth · Sales · Rescue · R&D · Special Ops]
- R & O & I & E & D & L & C & X -->|SITREP| HQ
- HQ --> IG[Inspector General]
- IG -->|verdict| HQ
- HQ --> H[History: lessons learned]
- HQ -->|final report| U
- ```
- ### Watch
- - [Stop Prompting Claude. Give It a Mission.](https://www.youtube.com/shorts/h9i1ZHay0JE) (YouTube Short)
- - [14 AI Units. 1 Command. (Claude Code Army)](https://www.youtube.com/shorts/2X-fzvEffys) (YouTube Short, recorded with v1.0's 14 units)
- ### See it
- | | |
- |---|---|
- | [](https://omer-yehuda.github.io/plugins/) | [](https://omer-yehuda.github.io/plugins/#units) |
- | **The mission:** a live operations table. HQ deploys units, then holds at the safety gate for your authorization | **Order of battle:** 18 units as shoulder patches; the rank stars follow the Lean/Full rations switch |
- | [](https://omer-yehuda.github.io/plugins/?unit=defense) | [](https://omer-yehuda.github.io/plugins/#roe) |
- | **Personnel files:** each unit's assignment, kit, doctrine and every field manual | **Rules of engagement:** safety gates, local memory and token rations, stated honestly |
- ### Highlights
- - **18 specialist units, 110 specialist playbooks:**
- - Research: competitor research, verification and wargames.
- - Code: bug hunting, unit and E2E testing, AI feature engineering.
- - Security: OWASP review, secrets scanning, hardening and AI security.
- - Legal: OSS licenses, privacy law, consumer law and contract review.
- - Text: UX copy and Hebrew/RTL localization.
- - Operations and market: CI/CD, SEO/ASO, data analysis, sales and support.
- - **Self-improving:** the R&D unit proposes new skills and playbooks from recurring lessons, and you approve every change.
- - **Cost-aware model routing:** a `lean` profile for the Pro plan (Sonnet and Haiku only) and a `full` profile for the Max plan (Opus for critical units). Switch with `/army:mode`.
- - **Token discipline by design:** units write full reports to disk and return SITREPs of at most 200 words. Recon stays fresh until manifests change, and field manuals load only on demand. Campaigns and Turbo show a dispatch preview before launching.
- - **Safety gates, bound to your mission:** a session-scoped `PreToolUse` hook tokenizes commands and asks before:
- - pushes, deletions, deploys and publishing
- - destructive SQL, `curl | sh` and downloaded packages
- - new dependencies, GitHub writes and external MCP actions
- - any attempt to switch itself off
- It's a speed bump, not a sandbox.
- - **Least-privilege units:** research and planning units can't run a shell or edit code. Every unit ignores instructions embedded in files and web pages, and reports them instead.
- - **Learns per project:** `.army/lessons.md` is read as data at the start of every mission, capped at 100 lines, and shown to you as a diff. It is never committed or packed, and a repository can't plant it.
- - **Turbo mode:** `raise the entire army` (or `all units on board`) engages every unit.
- See the full documentation in [plugins/army/README.md](plugins/army/README.md).
- ---
- ## Featured: Toon Shorts
- > One conversation, one finished cartoon short.
- Say **"make me an animated short"**. Claude asks a few questions, pitches 10 true-story ideas, writes a 30-60 second script built to hold attention, voices it, storyboards original toon characters on parallax sets, **checks its own preview frame by frame**, and renders the MP4 with a ready-to-paste title, description and hashtags.
- [](plugins/toon-shorts/README.md)
- - **Original characters:** 12 moods, 11 poses, 6 skin tones, and a wardrobe of hats, outfits and 28 props.
- - **20 sets with depth:** three parallax layers, 5 times of day, landmarks and foreground overlays.
- - **Karaoke captions** that light up the spoken word, with correct **Hebrew right-to-left** layout.
- - **Voice:** free Kokoro AI voices in 6 languages, your own recording in any language (timing snaps to your pauses), or captions only.
- - **Safe by construction:** nothing installs without asking, dependencies are pinned, and the model is SHA-256 verified. No telemetry.
- See the full documentation in [plugins/toon-shorts/README.md](plugins/toon-shorts/README.md).
- ---
- ## Privacy & safety
- - **No telemetry.** Plugins in this marketplace send nothing anywhere on their own.
- - **Local memory only.** Project memory lives in `.army/` inside your project, protected by its own `.gitignore` (`*`) and `.git/info/exclude`, so neither git nor `npm pack` includes it. Add `.army` to `.dockerignore` if you build images. Per-user settings live in Claude Code's plugin data directory.
- - **Human in the loop.** Irreversible actions require your approval by default.
- - **Toon Shorts installs nothing on its own.** Setup asks first, then installs into a private venv (`~/.cache/toon-shorts`) with pinned packages and a hash-verified voice model.
- - **Defensive security only.** The red-team specialist is restricted to your own project in a local or test environment.
- - The legal unit produces an automated risk review, **not legal advice**.
- ## Repository structure
- ```
- .claude-plugin/marketplace.json Marketplace catalog
- plugins/
- army/
- .claude-plugin/plugin.json Plugin manifest
- units.json Single source of truth: routing, tools, descriptions
- skills/ raise · raise-all · mode · gates · status · map
- agents/ 18 unit subagents (frontmatter generated from units.json)
- doctrine/ 110 field manuals, loaded on demand
- hooks/hooks.json Safety gate + SessionStart re-injection
- scripts/army.js Zero-dependency control script and gate
- scripts/build-map.js Generates the command-center map (fonts embedded)
- tests/ Gate + lifecycle tests
- CHANGELOG.md
- toon-shorts/
- .claude-plugin/plugin.json Plugin manifest
- skills/toon-shorts/ SKILL.md · references · examples · engine (scripts/toonshorts)
- tests/ pytest suite (coverage gate 85%)
- requirements*.txt Pinned dependencies
- CHANGELOG.md
- court/ customs/ er/ atc/ critics/ mission-control/ academy/ archive/ newsroom/ premiere/
- skills/ · agents/ · scripts/ (zero-dependency Node) · tests/ (node:test)
- skill-installer/
- skills/ install · browse
- scripts/si.js Zero-dependency installer (GitHub tree API + git fallback)
- NOTICE Credit: adapted from OpenAI's Apache-2.0 Codex skill-installer
- studio/
- skills/produce/ SKILL.md · examples · renderer (scripts/studiokit, Pillow + ffmpeg, Kokoro voice)
- skills/trend/ SKILL.md · trend desk: real footage, wide video + Short
- tests/ pytest suite
- docs/ Live site (GitHub Pages): interactive map + screenshots
- evals/ Each plugin vs. plain Claude on fixed missions
- templates/plugin-template/ Scaffold for new plugins
- tools/sync-army.js Generates agent frontmatter and tables, CI drift check
- .github/workflows/validate.yml CI: manifest validation + tests
- ```
- ## Contributing a new plugin
- 1. Scaffold it: `cp -r templates/plugin-template plugins/<name>`
- 2. Edit `plugins/<name>/.claude-plugin/plugin.json` (kebab-case name, description, version).
- 3. Add skills (`skills/<skill>/SKILL.md`), agents (`agents/*.md`) and hooks (`hooks/hooks.json`) as needed.
- 4. Register it in `.claude-plugin/marketplace.json`:
- ```json
- { "name": "<name>", "source": "./plugins/<name>", "description": "..." }
- ```
- 5. Validate it:
- ```bash
- claude plugin validate .
- claude plugin validate plugins/<name>
- ```
- 6. Try it locally: `claude --plugin-dir plugins/<name>`
- 7. Bump `version` in `plugin.json` for every release. Users stay on a pinned version until it changes.
- Plugin names must not start with `claude-` or `anthropic-`, because those prefixes are reserved.
- ## Development
- ```bash
- node --test plugins/army/tests/*.test.js # gate + lifecycle tests
- for p in court customs er atc critics skill-installer mission-control academy archive newsroom premiere; do node --test plugins/$p/tests/*.test.js; done
- (cd plugins/studio && python -m pytest) # studio renderer tests (pip install -r requirements-dev.txt)
- node tools/sync-army.js # regenerate files from units.json and standing orders
- node evals/run.js # plugins vs. plain Claude (needs a signed-in claude CLI)
- (cd plugins/toon-shorts && python -m pytest) # toon-shorts engine tests (pip install -r requirements-dev.txt)
- claude plugin validate . # validate the marketplace
- node plugins/army/scripts/build-map.js --out docs/index.html # rebuild the live map after changing the army
- ```
- CI runs all of the above on every push and pull request.
- ## License
- [MIT](LICENSE) © 2026 Omer Yehuda, except `plugins/skill-installer` (Apache-2.0, adapted from OpenAI's Codex skill-installer; see its NOTICE).
- Independent community project. Not affiliated with, sponsored or endorsed by Anthropic. Claude and Claude Code are trademarks of Anthropic, PBC.
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.