Six-stage WeChat Official Account content-publishing workflow with quality-gate tools and a bundled skill.
Install
# from GitHub (first run asks for allowBuilds approval — follow the hint, retry)
dsh plugin --profile web add github:Andiii208/gzhflow
Any plugin you install runs third-party code with your own permissions — it can read your files, use your credentials, and reach the network, and tool approvals don’t sandbox it. GitHub-sourced plugins also run build scripts at install time — pnpm blocks those until you allow them, so an install can stop with ERR_PNPM_GIT_DEP_PREPARE_NOT_ALLOWED or ERR_PNPM_IGNORED_BUILDS; dsh prints the exact key to add under allowBuilds in your profile’s pnpm-workspace.yaml, and the install works on the next run. Allowing a build is a trust decision: only install sources you trust, and pin a commit (github:owner/repo#sha).
README
A reusable, cross-agent workflow framework for publishing WeChat Official Account (公众号) content: from "a topic" to "a styled, illustrated, de-AI-flavored article in your draft box", as a six-stage pipeline with a quality gate at every step.
Works with Claude Code, Cursor, Codex CLI, Gemini CLI, Qwen Code, DeepSeek, and other mainstream agents. Also compatible with the Hermes
SKILL.mdformat (same-origin standard).
Features
- Cross-agent, out of the box:
AGENTS.mdas the single source of truth + a one-lineCLAUDE.mdbridge - Six-stage pipeline: material-first → writing → de-AI → illustration → layout → publish, each with a quality gate
- Personalization is pluggable: writing styles / image styles / layout themes live in
config/+examples/, never hard-coded into the flow - Three-tier quality gates: deterministic script gates + LLM self-check checklists + human review/dry-run
- Works for personal accounts: official API
draft/add(available to personal 订阅号) + manual paste fallback - Pure CLI tools: stdlib-first, no external dependencies, cross-platform
Quick Start
cp config/workflow.example.yaml config/workflow.yaml
cp config/styles.example.yaml config/styles.yaml
cp config/themes.example.yaml config/themes.yaml
cp config/publish.example.yaml config/publish.yaml
python scripts/validate_repo.py
# Then tell any agent:
# "Use the gzhflow workflow to write a WeChat OA article about <topic>"
See docs/quickstart.md for full instructions.
Note: WeChat OA publishing requires a Chinese WeChat Official Account and its AppID/AppSecret. Since 2025-07, personal accounts can no longer use the
freepublishAPI to publish directly —gzhflowpushes to the draft box viadraft/add(still available), and you tap "publish" in the 公众号助手 app yourself.
Use in DeepSeek Harness (DSH plugin)
gzhflow is also packaged as a DSH plugin (a dsh.bundle manifest in package.json):
# registers 7 gzhflow_* tools (6 quality gates + env self-check) and ships the gzhflow skill
dsh plugin --profile web add https://github.com/Andiii208/gzhflow
Then tell any session "Use the gzhflow workflow to write a WeChat OA article about ", or load the gzhflow skill.
For the "公众号主编" persona (session-scoped, opt-in), install the agent preset:
powershell -ExecutionPolicy Bypass -File dsh-preset/install.ps1 # Windows
bash dsh-preset/install.sh # macOS / Linux
See dsh-preset/README.md.
Pipeline
| Stage | Purpose | Gate |
|---|---|---|
| ① Material-first | Ask questions to mine real material — never fabricate | Intent triple-check |
| ② Writing | Write in the routed style | ai_flavor_score.py + self-check |
| ③ De-AI | Structural surgery + voice + person adaptation | ai_flavor_score.py re-check |
| ④ Illustration (opt.) | Source triage → style routing → 4-part prompt | check_image_prompt.py |
| ⑤ Layout | Theme routing → Markdown → WeChat HTML | validate_gzh_html.py |
| ⑥ Publish | Push to draft box | publish_draft.py --dry-run |
License
Links
More in this category
Q00/ouroboros#integrations/dsh-plugin★ 5768
Config-only bundle that mounts Ouroboros through the DSH MCP client, exposing 36 interview, Seed, execution, evaluation, and evolution workflow tools in DSH.
huangruiteng/loopx#dsh-loopx-plugin★ 5584
LoopX, a provider-neutral, local-first state kernel and control plane for long-horizon agents: keeps Goal, Todo, gate, evidence, quota, recovery, and handoff state above DeepSeek Harness, while the plugin bootstraps the CLI and skills, admits bounded same-session continuation, and adds a loopback GoalBar for the exact bound loop.
chuspeeism/dashi-taskboard#deepseek-harness★ 2914
Embeds the active installed Codex Taskboard runtime in the DeepSeek Harness sidebar, using its launcher runtime descriptor instead of a fixed port.
NanmiCoder/dsh-agent-teams★ 1345
AgentTeams multi-agent teams.
tong-io/tongflow#dsh-tongflow★ 1014
TongFlow film-crew studio for image, voice, music and video production: the agent writes per-asset TongFlow workflow files (.tongflow.json) that run through TongFlow plugins, with an embedded workflow canvas, a shot/character/take project layout and a manga-drama template; sessions starting with @tongflow open the Studio view.
sandbaseai/sandbase-harness★ 642
Connects dsh to a local SandBase Harness runtime over stdio MCP for managing agents and sessions, streaming turns, inspecting artifacts and cancelling work.
Community comments
Comments are public GitHub Discussions. Loading them connects to GitHub and Giscus; a GitHub account is required to post.