plugin-development
Skill33 starsChanged 22 days agoMITView on GitHub
Use when packaging skills, commands, hooks, and MCP servers into a distributable plugin. Covers manifest structure, bundling, versioning, testing, and distribution through a marketplace. Trigger on "package my skills as a plugin", "publish to a plugin marketplace", "share our commands with the team", "bump the plugin version". Do not use for writing one skill — see the skill-authoring skill; for building the MCP server itself, see the mcp-server skill.
We split this skill into the rules it gives an agent
2 rules. 0 are common, 2 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
2
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 57
Do not bundle a skill that duplicates something the agent does well already.
- Distinctive: line 58
Test the uninstall path. A plugin that leaves hooks behind after removal is a bug that is hard for a user to diagnose.
“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Plugin Development
Purpose
Package a coherent set of agent capabilities — skills, slash commands, hooks, and MCP servers — into something another person can install in one step and that works immediately.
When to Use
- Distributing skills or commands to a team or publicly.
- Bundling a set of related capabilities that belong together.
- Publishing to a plugin marketplace.
- Versioning and updating an existing plugin.
Capabilities
- Plugin structure and manifest.
- Bundling skills, commands, hooks, agents, and MCP servers.
- Versioning and compatibility.
- Testing a plugin before distribution.
- Marketplace publication.
Inputs
- The capabilities to bundle, and their coherence as a set.
- The audience and what they already have installed.
- Any external dependencies: MCP servers, API keys, binaries.
Outputs
- A plugin that installs cleanly and works without further configuration.
- A manifest with accurate metadata.
- Documentation of what it provides and what it requires.
Workflow
- Bundle around a coherent purpose — A plugin is a set of capabilities that a specific kind of user needs together. A grab-bag of unrelated skills is a worse experience than several focused plugins.
- Write the manifest accurately — Name, description, version, and what it provides. This is what users see when deciding whether to install.
- Declare the dependencies honestly — If a skill requires an MCP server, an API key, or a binary on the PATH, say so prominently. A plugin that silently fails on a missing dependency will be uninstalled.
- Test from a clean install — In an environment without your local configuration. Plugins that "work on my machine" are the standard failure.
- Version semantically — A breaking change to a skill's behavior is a major version. Users have workflows built on it.
- Document what it does, not what it is — Users care about what problems it solves.
Best Practices
- A plugin's skills should not overlap with each other. If two skills in one plugin trigger on the same request, they will both load and cost double.
- Declare external requirements at the top of the README. The most common plugin failure is an unstated dependency.
- Provide a working example in the README — a real request and what the plugin does with it. This is worth more than a feature list.
- Do not bundle a skill that duplicates something the agent does well already. It adds cost and no capability.
- Test the uninstall path. A plugin that leaves hooks behind after removal is a bug that is hard for a user to diagnose.
- Pin any MCP server version the plugin depends on. An upstream change to a tool's schema will break your skills silently.
Examples
Plugin structure:
my-plugin/
.claude-plugin/
plugin.json the manifest
skills/
review-pr/SKILL.md
triage-issue/SKILL.md
commands/
standup.md a slash command
hooks/
hooks.json format-on-write
.mcp.json the MCP servers this plugin needs
README.md{
"name": "engineering-workflow",
"version": "1.2.0",
"description": "Pull-request review, issue triage, and standup summaries for teams working in GitHub and Linear.",
"author": { "name": "Nima Dorostkar" },
"homepage": "https://github.com/nimadorostkar/claude-skills",
"keywords": ["code-review", "triage", "github", "linear"]
}A README that prevents the most common failure:
## Requirements
This plugin requires two MCP servers, which are configured automatically on
install but need authorization:
- **GitHub** — run `/mcp` and authorize. Without it, `review-pr` cannot read diffs.
- **Linear** — run `/mcp` and authorize. Without it, `triage-issue` will fail
with "no Linear workspace".
The plugin will install successfully without these, but the skills will not work.
## Example
> review the PR at github.com/acme/api/pull/412
The `review-pr` skill fetches the diff, reviews it for correctness, security,
and performance, and posts findings grouped by severity as review comments.Notes
- The most common plugin defect is an undeclared dependency: the author has an MCP server configured globally, so the plugin works for them and fails for everyone else. Always test in a clean profile.
- Marketplace descriptions are the plugin's only chance to be found. Include the tools and the tasks, in the words a user would search for.
- Breaking a skill's behavior without a major version bump breaks workflows built on it. Skills are an interface.
Other versions4 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 4
- jokob-sk/NetAlertX★ 7.3ksame name
- netalertx/NetAlertX★ 7.3ksame name
- matrixorigin/Memoria★ 621same name
- prismatic7/agy-plugin-development★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository137 other files it gives its agents.
- agent-instructionsSkill
- agent-memorySkill
- hooksSkill
- skill-authoringSkill
- skill-reviewSkill
- slash-commandsSkill
- subagentsSkill
- agent-designSkill
- context-engineeringSkill
- fine-tuningSkill
- llm-cost-optimizationSkill
- llm-evaluationSkill
- llm-integrationSkill
- mcp-serverSkill
- ml-pipelineSkill
- model-selectionSkill
- prompt-engineeringSkill
- ragSkill
- structured-outputSkill
- api-designSkill
- background-jobsSkill
- cachingSkill
- djangoSkill
- event-driven-architectureSkill
- fastapiSkill
- graphqlSkill
- microservicesSkill
- nestjsSkill
- realtime-websocketsSkill
- spring-bootSkill
- competitive-analysisSkill
- lead-researchSkill
- product-analysisSkill
- resumeSkill
- user-researchSkill
- database-performanceSkill
- data-modelingSkill
- data-qualitySkill
- pandasSkill
- postgresSkill
- sparkSkill
- brand-guidelinesSkill
- generative-artSkill
- themingSkill
- ui-reviewSkill
- visual-designSkill
- architecture-decisionsSkill
- bug-fix-protocolSkill
- cli-developmentSkill
- code-reviewSkill
- debuggingSkill
- design-patternsSkill
- domain-modelingSkill
- git-workflowSkill
- legacy-modernizationSkill
- refactoringSkill
- repository-explorationSkill
- simple-designSkill
- system-designSkill
- aws-cdkSkill
- The first 60 shown.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

