cornus

Audit the licenses of every third-party dependency Cornus actually ships (the Go modules compiled into the binaries plus the npm packages bundled into the embedded web SPA), classify each as permissive / weak-copyleft / strong-copyleft, flag anything that needs review, and regenerate THIRD_PARTY_NOTICES.md and the NOTICE stanza. Use when adding or bumping dependencies, before a release, or whenever you want to re-verify license compliance.

Read `.agents/docs/JOURNAL.md` and `.agents/docs/LTM/` documents, find durable knowledge that belongs in the canonical docs, and update those documents with concise synthesized findings. Targets are `.agents/docs/OVERVIEW.md`, the root `ARCHITECTURE.md`, and the user-facing VitePress site under `docs/`. Use when journal or LTM notes have accumulated implementation, system, or user-facing knowledge that should be promoted into the project's canonical overview, architecture, or user reference docs.

Read JOURNAL.md and reorganize its contents into semantically separate long-term memory documents under .agents/docs/LTM/. Use when you want to consolidate accumulated journal entries into reusable, topic-organised reference files.

Read TODO.md and scan source code for TODO/FIXME comments, build a consolidated list, then dispatch parallel agents to address as many items as possible.

Translate, localize, review, or synchronize Markdown documentation while preserving source meaning, document structure, code, configuration, links, and project terminology. Use for creating or updating locale trees such as docs/ja or docs/zh, auditing an existing translation against its source, fixing mixed-language or unnatural translated prose, or propagating source-document changes into translations.

Capture a screenshot of a web page — primarily the `cornus web` SPA running locally (localhost:5000), or any URL — and embed it into a document (VitePress docs/, .agents/docs notes, or generic markdown) with the image saved to the correct asset location and a proper markdown/HTML embed. Use when asked to screenshot the web UI or a page for docs, add a UI image to a guide, or illustrate a document with a captured web view.