
md2wechat 3.x for Agents: What Changed
From the 3.0 single-source layout catalog through 3.2 preflight and the official OpenClaw installer to 3.4 subject-reference images.
md2wechat 3.x for Agents: What Changed
The 3.x line is about determinism for agents: one catalog, preflight before side effects, and official installers. Layout catalog.
3.0 — the catalog became the single source
User layout YAML overrides were removed. The embedded catalog (56 recommended syntax names, 77 scenario entries, 63 render-layer capabilities) is the only source of truth, and unknown fields or variables fail closed. For agents this means discovery output is stable across machines.
3.2 — preflight and the official OpenClaw installer
Preflight moved intent, asset, and cover validation before any auth or remote call — a failed run no longer burns a credential check. OpenClaw distribution switched to the official npm-based metadata.openclaw installer (openclaw skills install @geekjourneyx/md2wechat).
3.3 — the counts contract
The layout catalog counts (56 / 77 / 63) became a pinned contract with conformance witnesses — 56 canonical plus variants and 3 compatibility modules. Docs and discovery now agree by construction.
3.4 — subject-reference images
generate_image --subject-reference <https URL> keeps a consistent subject across generated images, on MiniMax image-01 only, with supports_subject_reference exposed in providers list. The publishing-oriented flow pairs with plan mode for host-agent rendering (Image Generation).
Verify what you run
md2wechat version --json
md2wechat capabilities --jsonVersion facts above are the changelog record; verify your installed binary before wiring a pipeline to any of it.
More Posts

What a WeChat Automation Workflow Should Include
A workflow breakdown covering content input, formatting, drafts, media, and review in a WeChat publishing pipeline.

How to Design an Agent Content Topic Selection Workflow
A practical guide to AI Agent topic selection workflows: why topic judgment matters more than drafting, what a topic workflow should output, how jina-cli supports source retrieval, and how the result connects to md2wechat publishing.

How to Turn AI Sources into an Automated Writing Pipeline
A workflow guide to AI source automation, covering fetching, filtering, deduplication, validation, and topic-card generation for writing systems.
Newsletter
Join the community
Subscribe to our newsletter for the latest news and updates