Canonical manifest schema
plugin.json targets the official 1.0.0 schema.
Agent Plugins specification 搂5.2
Create and improve Agent Skills through intent capture, focused SKILL.md authoring, validation, test prompts, qualitative review, benchmarks, and iterative refinement.
measured
then improved.
Intent becomes instructions; instructions meet test prompts; evidence drives the next revision.
Sparse-clone the plugin directory and load its root in a compatible client. Review the bundled Python evaluation and packaging scripts before running them.
Installation
Install Plugin
Choose Codex or Copilot, click Copy, paste the two lines into a terminal, and press Enter.The installation commands are identical on Windows, macOS, and Linux.
codex plugin marketplace add agentplugin-net/portable-agent-plugins
codex plugin add skill-creator@agentplugin-netExplainable standards check
Official latest version: 1.0.0 (Published). The value is read from the official specification and cached for one day.
This is a static package-format check, not a security audit, quality score, malware scan, or guarantee of client compatibility.
plugin.json targets the official 1.0.0 schema.
Agent Plugins specification 搂5.2
The plugin name 鈥?{name}鈥?satisfies the 1鈥?4 character naming rules.
Agent Plugins specification 搂搂5.3 and 5.5
No unknown top-level manifest fields were found.
Agent Plugins specification 搂5.2
All present metadata and extension fields use permitted JSON types.
Agent Plugins specification 搂搂5.2鈥?.6
A regular plugin.json file exists at the plugin root.
Agent Plugins specification 搂搂4.1 and 5.1
1 immediate-child Agent Skill discovered; every SKILL.md passed required frontmatter and naming checks.
Agent Plugins specification 搂7.1 and Agent Skills specification
No root mcp.json is present; MCP is optional and its absence is conforming.
Agent Plugins specification 搂搂6.2 and 7.2
Capabilities
Treat Skill authoring as an iterative product process.
Use this plugin when creating a new Skill or when an existing Skill triggers poorly, produces uneven work, or lacks evidence that revisions help.
The name and description are treated as routing interfaces, while the body stays focused on the workflow and reusable resources.
Test prompts, graders, variance analysis, and review artifacts make improvements inspectable and reveal overfitting risks.
Practical workflow
Clarify the task, triggers, tools, edge cases, and reusable assets, then write the smallest effective Skill.
Run representative prompts, review outputs and scores, fix observed problems, and repeat with a broader test set.
Package contents
Skill Creator declares 1 Skill, 0 MCP configurations, and 3 additional resources. The included-file links below point to the reviewed source repository.
Agent Skills
1
skill-creator
MCP servers
0
No runtime service
Package format
Agent Plugins
Official latest 1.0.0
Synced source content
This README snapshot is stored by Agent Plugins Hub and refreshed by the daily GitHub synchronization pipeline. Remote images and embedded content are not loaded.
This repository does not currently provide a README snapshot. The plugin instructions below remain available.
Source documentation
Rendered from the linked GitHub source. Raw HTML is disabled, scripts cannot execute, remote images are omitted, and external links open separately.
Skill Creator supports the complete Agent Skill lifecycle: intent capture, SKILL.md authoring, validation, test-prompt design, qualitative review, benchmarking, description improvement, packaging, and iterative refinement.
Define what the Skill should do and when it should trigger. Write the smallest effective instructions and reusable resources, then test representative positive, negative, and edge prompts. Compare results, fix observed failures, and repeat with a broader test set.
Benchmark results depend on prompts, graders, models, and sampling variance. Do not treat a score as universal quality. Review bundled Python scripts before running them and avoid sensitive production data in evaluation fixtures.
This portable package is maintained in agentplugin-net/portable-agent-plugins from a pinned Apache-2.0 upstream Skill.
Continue exploring
A portable Agent Plugin for retrieving current, version-specific library documentation and code examples through Context7 MCP.
An Agent Plugin package connecting MCP clients to Exa for web search, code search, content fetching, crawling, and multi-step research.
Create, connect, and interact with an AlloyDB for PostgreSQL database and data.