📝 Add long-form editorial article template/profile paired with the Medium-style Copilot agent #379

October 29, 2026 — editorial implementation in draft PR #468

Draft PR #468 implements the article profile on top of merged #467. Current head: 256b0e88884db3d6dd768d7acca89d143b8f119d; tree: aab15a096c9b57ef5dd063637adb268ad06e5098. Uploaded blobs and complete trees match the local commits.

The existing renderflow.profile/v1/ordinary-planner/strategy path now has a bundled editorial selection and editorial-v1 presentation. It declares HTML/PDF, optional DOCX/EPUB and an exact original Markdown copy. Bundled styles, templates, the Pandoc article filter, shared LaTeX components and digest-pinned licensed fonts are embedded in the binary. YAML/CLI selection and SDK planning use the same profile contract.

The article guide defines renderflow.editorial/v1 for #378. The synthetic example exercises metadata, hero alt/caption, headings, quotations, tables, code, notes, images and footnotes. No writing agent or AI runtime is added.

Pandoc >=3.6.1 is required; PDF uses cached-only Tectonic. External source images and user font registries are not snapshotted at planning, so editorial render cache/checkpoint reuse is disabled; observed font drift is rejected before rendering. HTML/PDF styling, DOCX/reference-font differences, EPUB reflow, English labels and regular-font/math/script limitations are documented explicitly.

Evidence is source review only. Nine regression cases are authored and unrun, including three explicit native opt-ins. Only the synthetic PNG authoring recipe and source asset/hash operations were executed. Tests, compiler/build/Cargo, lint/format/schema/docs checks, product CLI/native/provider execution and CI dispatch/polling remain deferred. No rendered or visually verified output is claimed; no dependency, lockfile, workflow, tag or release change was made.

This issue stays open with its acceptance criteria preserved until executable/visual/accessibility/reproducibility evidence exists. Handoff.

Description

Add a reusable long-form editorial article template/profile that pairs with the repository's Medium-style Copilot agent.

The purpose is not to reproduce Medium's proprietary visual design. The target is a clean, readable, publication-quality long-form article experience inspired by modern editorial web writing while remaining distinctly Renderflow and provider-neutral.

This issue should inspect Renderflow's current template/profile architecture before implementation. Use the current canonical publication/profile abstractions rather than reviving an older or superseded template path merely because early Renderflow prototypes used one.

Goals

• provide a first-class long-form article presentation profile
• pair content generation/polishing with deterministic rendering
• support local-first, reproducible assets and typography
• render cleanly across the output formats that the current capability graph can truthfully support
• serve as a reusable fixture/example for author → polish → render workflows

Relationship to Copilot Agent

The paired agent is responsible for content:

.github/agents/medium-article.agent.md

The template/profile is responsible for presentation.

Maintain this boundary:

rough notes / draft
  ↓
Copilot generation + polish
  ↓
canonical article Markdown + metadata
  ↓
Renderflow editorial profile
  ↓
validated derivatives

The template/profile must not depend on Copilot at runtime.

Content / Metadata Contract

Support, where available in the current Renderflow profile system:

• title
• optional subtitle/dek
• author/contributor metadata
• publication/update date
• canonical URL metadata when explicitly configured
• tags/topics/keywords
• hero/cover image
• inline images and captions
• code blocks
• block quotations
• callouts or asides
• links and footnotes
• accessibility/alt-text metadata
• optional table of contents for sufficiently long articles

Do not invent missing metadata.

Presentation Requirements

Aim for a restrained, readable editorial system with:

• strong typographic hierarchy
• comfortable measure and line height
• readable body typography
• clean spacing between sections
• high-quality code and quotation presentation
• responsive HTML behavior where HTML is supported
• sensible print/PDF pagination where PDF is supported
• graceful representation in DOCX/EPUB/editorial interchange outputs where current adapters support them

Avoid hard-coding presentation details into source Markdown when they belong in the profile.

Local-First Assets

Use repository-owned or explicitly licensed local assets wherever practical.

Do not require runtime calls to Google Fonts, CDNs, or other network resources for the canonical deterministic build.

If the profile uses fonts, stylesheets, icons, or other assets:

• keep provenance/license information
• pin the exact asset revision or bytes
• integrate with Renderflow's current asset/toolchain model
• provide deterministic fallbacks

Multi-Format Behavior

Inspect the current format capability graph and implement only outputs that are genuinely supported.

At minimum, the profile should be designed to support the article semantics across:

• HTML/web article output
• PDF/printable article output

Where current capabilities permit, also validate:

• DOCX editorial interchange
• EPUB/e-book derivative
• normalized Markdown/source derivative

Do not claim parity where an output format necessarily loses presentation semantics. Document the expected fidelity/loss profile.

Example / Fixture

Add a redistribution-safe synthetic article fixture that exercises representative features such as:

• metadata
• headings
• prose
• links
• quotation
• code block
• image/caption
• list
• callout/aside if supported

Use it for deterministic rendering tests and documentation examples.

Validation

Validate the article profile through the repository's current validation/conformance mechanisms.

Check, where applicable:

• template/profile discovery
• required local assets
• typography/font availability
• missing metadata behavior
• HTML structure
• PDF generation
• cross-format degradation
• reproducibility
• accessibility-relevant semantics

Naming / Trademark Constraint

Use "Medium-style" only as workflow/design inspiration in internal context where useful. Public-facing naming should prefer a neutral term such as editorial, long-form, or article unless there is a specific reason otherwise.

Do not copy Medium branding, proprietary CSS, logos, layouts, or assets.

Acceptance Criteria

• Current Renderflow template/profile architecture is inspected before implementation
• A reusable long-form editorial/article profile is implemented using the canonical architecture
• The profile's input metadata/content contract aligns with the Medium-style Copilot agent
• HTML output is clean, responsive, and readable when supported
• PDF output is clean and publication-quality when supported
• Additional supported derivatives degrade honestly and predictably
• Canonical builds require no runtime font/CDN downloads
• Asset/font provenance and deterministic fallbacks are documented
• A redistribution-safe synthetic article fixture is included
• Representative outputs are covered by tests/validation
• Documentation explains generation/polishing → rendering workflow
• Formatting, linting, tests, and CI pass

Audit-to-roadmap reconciliation — 2026-10-21

Audit: https://github.com/egohygiene/renderflow/blob/audit/flow-suite-20261021T025401Z/audits/flow-suite-20261021T025401Z.md · Complete roadmap: #367

Implement against current profiles; local licensed assets, honest cross-format degradation.

Bounded preparation: Inspect existing canonical profile path and align metadata contract with #378.

Remaining acceptance (common qualification deferral; no dependency star): Rendered fixture, typography/accessibility and deferred build/test/CI qualification.

The maintainer's implementation-sprint qualification deferral remains in force. This audit ran no Renderflow tests, compiler/build/Cargo resolution, product lint/format/schema qualification, native/CLI/provider/visual execution, installers or new CI. Reading existing exact-source provider records is not a qualification rerun. Source implementation and authored tests are not passing-test, native-fidelity, release or downstream-consumer evidence.

This is an additive planning note. The full original issue, acceptance criteria, scope, parent references and historical checkpoints remain authoritative and unchanged. No acceptance box or issue is closed by this audit. Recommended ledger parent: #367.