メニュー
Reyn-specific usage cheat sheet -- which mechanism to reach for (skill/pipeline/mcp/hook/present), composition idioms, op essentials, and pointers to the full specs. Read this before authoring a new part or composing several.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Make a folder of the operator's own documents (txt/md/pdf/xlsx/pptx/docx) searchable by meaning via a user-named sqlite vector store. Covers installing the `rag` plugin, embedding-provider setup, the exact `rag_ingest`/`rag_query` pipeline calls, and corpus internals -- via bundled references. Read this before running the builtin `rag_ingest` / `rag_query` pipelines, or when the operator asks you to search a folder/file of their own documents.
How to build something that REACTS to an external system (an orchestrator, a watcher, a UI, any MCP server that pushes) -- which reyn mechanism already answers each requirement, and the anti-pattern list of things people re-invent. Read this BEFORE designing any external-event-driven plugin, server-push handling, wake/notification behaviour, or browser UI integration. Companion to reyn_cheat_sheet (which covers choosing between skill/pipeline/mcp/hook/present in general).
Draft an artifact, self-review it against your own checklist via a schema-validated agent step, and revise on failure -- the standard Evaluation-gated workflow for any "produce then check quality" task (a summary, a doc section, an email, a report paragraph). Read this before handing off a self-authored artifact you have not gated.
Translate an English document into Japanese, preserving headings, lists, and code blocks.
A minimal test skill that includes search_hints for FP-0024 Component B invariant tests.
| name | reyn_cheat_sheet |
| description | Reyn-specific usage cheat sheet -- which mechanism to reach for (skill/pipeline/mcp/hook/present), composition idioms, op essentials, and pointers to the full specs. Read this before authoring a new part or composing several. |
This is the gap-filler between "reyn has these parts" and "you use them correctly". concept/reference docs describe each mechanism in isolation; this skill is the composition know-how in the cracks between them (0060 Addendum D1). Read on demand when deciding which mechanism to use, or before authoring a new skill/pipeline/hook/present-view.
This L2 router stays deliberately small: it is enough on its own to decide
whether you need to go deeper and which references/ file answers your
question -- you should not need to open a reference just to find out if you
need it.
hook | mcp |
the builtin build_and_query_rag_corpus skill (document search over the
operator's own files). Deeper detail (hook action schemes, a
CI-verified worked hook example, MCP tool discovery):
input-hooks-and-mcp.md
-- read when wiring a reactive trigger or calling an external tool/resource.skill | pipeline | an
mcp tool call mid-flow. Deeper detail (pipeline step grammar, the
flagship CI-verified worked pipeline, SKILL.md authoring shape):
workflow-pipelines-and-skills.md
-- read when writing pipeline DSL steps or authoring a new skill.present |
render_template | an mcp write. Deeper detail (present's blueprint
catalog and the input/output caveat, the self-review gate before
promotion):
output-present-and-self-review.md
-- read when deciding whether/how to show a result via present, or
self-reviewing an authored asset before it becomes a reused part.Reuse before authoring: check the existing catalog (list_actions) for a
part that already covers the need. Author only when nothing fits, and
self-review anything you author or promote (an agent step + schema, see
output-present-and-self-review.md)
before it becomes a reused asset -- an ungated authored part is a liability,
not a shortcut.