wechat-comic-factory

# Wechat Comic Factory Execute local commands first. Do not replace execution with creative brainstorming, platform advice, prompt collections, or simulated publish confirmations. ## Execute Only Local Entrypoints Use only these public entrypoints: ```powershell python scripts/run_pipeline.py --comic_type "小林漫画" --topic "成年人的体面" --count 3 python scripts/run_pipeline.py --comic_type "育儿漫画" --topic "当妈三年我学会了精打细算" --count 4 python scripts/run_pipeline.py --publish_latest python scripts/run_pipeline.py --publish_output_dir "<existing_output_dir>" ``` Do not expose底层多脚本串联命令给用户，不要跳过 `run_pipeline.py` 直接拼接发布流程。 ## Config Guardrail If the local script returns `stage = config`, or the error message indicates `config.json` is missing / still placeholder: 1. Stop execution reporting immediately. 2. Tell the user to create `config.json` from `config.example.json`. 3. Ask them to fill real `text_llm`、`image_llm`、`wechat` credentials before retrying. Use a short, direct response such as: - `当前还没有可用配置。请先在技能根目录把 config.example.json 复制为 config.json，并填入真实模型与公众号配置。完成后我再执行。` Do not fabricate partial success, draft status, or publish status when configuration is incomplete. ## Generation Rules When the user asks to generate comics: 1. Parse `comic_type`, `topic`, and `count`. 2. If `count` is missing, default to `3`. 3. If `comic_type` is missing, ask for the type. 4. If `comic_type` is unsupported, report the supported types directly. 5. Execute `python scripts/run_pipeline.py ...`. 6. Return the script JSON result instead of writing creative copy first. Supported types: - `小林漫画` - `育儿漫画` Never rewrite the user’s topic unless they explicitly ask for rewriting. ## Publish Rules In this skill, “发布” means only: - push to **微信公众号草稿箱** It must never mean: - personal WeChat chat - WeChat group - current chat reply - message tool sending - any other external channel When the publish target is clear: 1. Use `python scripts/run_pipeline.py --publish_latest` for the latest pending task. 2. Use `python scripts/run_pipeline.py --publish_output_dir "<path>"` only when a specific task directory is explicitly identified. ## Clarify Ambiguous Follow-ups Clarify instead of guessing when the user says: - `发布这个` - `直接发` - `发出去` - `1，2直接发布` - `直接发微信` Recommended clarification style: - `我当前只支持推送到微信公众号草稿箱。你是要发布最近一次生成的整组内容吗？` - `最近任务是《成年人的体面，藏在细节里》三张系列。是否推送该整组内容到微信公众号草稿箱？` - `我暂不支持只发布前两张。你是要发布整组内容，还是先重新生成一组两张的版本？` Do not ask about contacts, group chats, personal accounts, GitHub, blog platforms, or other channels. ## Success Rules You may say: - `已发布` - `已推送到公众号草稿箱` - `草稿已创建成功` only when the script JSON result includes both: - `success = true` - `draft_status = success` Otherwise say only: - `已生成待发布内容` - `尚未发布` - `发布失败，失败阶段为 xxx` - `未检测到可用微信草稿发布结果` If the script fails, report the real `stage` and `error` fields. Never convert a failed script run into a success message.