菜单
Use when creating or structuring a Claude Code plugin for this marketplace — the on-disk layout, the plugin.json manifest fields, the frontmatter for skills/commands/agents, and the authoring rules (no absolute paths, least-privilege tools, one coherent capability).
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Use when designing memory or persistent context for an agent — deciding what to persist vs. recompute, how to retrieve only what's relevant, and how to summarize/compact history to fit the context window. Use when an agent forgets, bloats its context, or repeats work.
Interactive installer for Rosetta. Guides users through selecting and installing skills and language-specific rules at user-level or project-level, verifies paths, and optionally optimizes installed files.
Instinct-based learning system that observes sessions via hooks, creates atomic instincts with confidence scoring, and evolves them into skills, commands, and agents.
Use when building or reviewing an evaluation for an LLM feature — assembling a representative test set, choosing pass criteria (exact match, programmatic checks, rubric, or LLM-as-judge), and catching regressions. Use when asking "how do I know this prompt or model change is better?"
Use when writing or reviewing a prompt for an LLM — making it specific, structured, and testable (clear task, role/context, examples, an explicit output contract, decomposition, room to reason) and avoiding the anti-patterns that make prompts brittle. Provider-agnostic.
Pattern for progressively refining context retrieval to solve the subagent context problem.
| name | plugin-authoring |
| description | Use when creating or structuring a Claude Code plugin for this marketplace — the on-disk layout, the plugin.json manifest fields, the frontmatter for skills/commands/agents, and the authoring rules (no absolute paths, least-privilege tools, one coherent capability). |
A plugin is one coherent capability authored once in Claude Code's format. Everything else (Cursor, Codex, OpenCode, Gemini, Copilot) is generated from it by the adapters, so author for the canonical format and let translation handle the rest.
my-plugin/
├── .claude-plugin/plugin.json # required manifest
├── README.md # required: what / why / how-to-use
├── skills/<name>/SKILL.md # optional: model-invoked knowledge
├── commands/<name>.md # optional: user-invoked slash commands
├── agents/<name>.md # optional: subagents
├── hooks/hooks.json # optional (Claude Code only — note it in the README)
└── .mcp.json # optional: MCP servers
Directories are auto-discovered; presence on disk is registration.
plugin.jsonRequired: name (kebab-case, unique), description (one sentence), version (SemVer),
author.name, license (SPDX). Recommended: keywords (drives search). Category is not a
manifest field — it lives catalog-side in marketplace.json.
name + description. The description is the trigger: write "Use when …".description, optional argument-hint, and least-privilege allowed-tools.name + description (when to delegate), optional tools and model.CLAUDE_PLUGIN_ROOT variable instead of hard-coding a path. (Claude Code interpolates that
variable wherever it literally appears in a loaded skill or command — so name it in prose,
as here, rather than writing the ${…} token, or your text renders as an absolute path.)allowed-tools / tools narrowly — it feeds both security review
and adapter fidelity.writing-effective-skills skill).Prefer skills and commands — they translate cleanly everywhere. SKILL.md is now a
cross-harness open standard, so skills are the highest-leverage, most portable component.
Treat hooks as Claude-Code-only enhancements, never load-bearing behavior.