| name | project-output-article |
| description | Canonical structure and LLM-first table of contents for long-form project output articles under projects/<name>/outputs/*.md (system design, interconnection, behaviour). Pair with sysml-view-doc-sync for deploy alignment; mermaid + mmdc for diagrams. Triggers: outputs article, system design doc, project markdown standard, llm_toc, llm_keywords, outputs TOC, foam detection style doc. |
| metadata | {"pattern":"pipeline","pairs_with":["sysml-view-doc-sync","system-design-report-generator","mermaid","mmdc","md-to-tex","skill-creator"]} |
| token_guardrails | - Normative article rules live in references/PROJECT_OUTPUT_ARTICLE_STANDARD.md — do not fork a second source of truth under docs/.
- Model/deploy is authoritative; article text and Mermaid labels must match deploy part and port paths (de facto wiring).
- These outputs are primarily for **human reading**: keep prose concise, precise, and concrete. Prefer tables, bullets, and short captions over long narrative blocks.
- Avoid repeating the same fact in multiple places. Say it once in the clearest section, then reference it.
- Do not paraphrase the model with extra adjectives or filler; if the model already expresses a detail cleanly, mirror it tersely or omit it.
- Prefer one top **nav fence** (YAML or JSON) immediately after title + source: `llm_toc` plus optional `llm_keywords` in the **same** block; keep keyword map small (~15); optional human markdown TOC below. For token efficiency use **partial reads** (first fenced block at top, or grep `llm_toc` / `llm_keywords`) and **`#id` anchors** — do **not** bake in line numbers (they change when the file edits).
- Mermaid: embed in .md; validate with mermaid skill; use mmdc only when rendering assets is requested.
- Submodule / canonical repo: [project-in-another-repo.mdc](../../../.cursor/rules/project-in-another-repo.mdc). Scope: [AGENTS.md](../../../AGENTS.md).
|
Project output article
When: Authoring or restructuring a long projects/<name>/outputs/*.md article (system design, PAT doc, interconnection) so humans and agents share one section order and navigable outline.
These articles are usually read by people first. Favor compact section prose, crisp headings, and traceable tables over model-echoing paragraphs.
Pairing
- sysml-view-doc-sync — content must match
.sysml deploy/behaviour; apply that skill for diffs and port paths.
- system-design-report-generator — folder + section
.md files + hub llm_toc with file when the report is split for LLM partial reads; still follow this standard’s section semantics.
- mermaid / mmdc — diagram text vs CLI render.
- md-to-tex — optional export to one
.tex from one or several outputs/*.md (pandoc multi-input or merge-first; use llm_toc order when bundling); not a substitute for article structure rules here.
- skill-creator — only when promoting this pattern into another repo’s skill pack (portability).
Pipeline
- Read the standard — references/PROJECT_OUTPUT_ARTICLE_STANDARD.md (section order,
llm_toc, optional llm_keywords, Mermaid rules).
- Choose shape — Monolith: start from assets/article-shell.md. Multi-file pack: system-design-report-generator (
outputs/system-design-report/, hub + sections).
- Multi-file packs: Section files are inputs to merge, NOT standalone documents. ("See section X", "See chapter Y") before final merge. Replace with (same-file anchor links).