Ls xhs note

Skill FrankL1u/liusir-skills/skills/ls-xhs-note

AI agent skills for WeChat article writing, publishing, and workflow automation.

Install

npx -y skills add FrankL1u/liusir-skills --skill ls-xhs-note

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Use when the user wants a complete Xiaohongshu note workflow that can start from a topic and turn it into a source article, a native note, and image assets, or continue from an existing note draft. Supports framework selection, model-native note drafting, and image generation.

SKILL.md

6.5 KB, ~1.2k tokens by cl100k_base, as published. Nobody here has run it

XHS Note Workflow

Create Xiaohongshu note assets from a topic or an existing draft.
Workflow rules live in references/. Repeatable execution lives in toolkit/. Content decisions belong to the model and the reference files; the CLI stays on the execution side.

Setup

Run cd toolkit && npm install && npm run build.
Run pip install -r requirements.txt.
Choose a runtime root: project-local ./.ls-xhs-note/ or user-level ~/.liusir-skills/ls-xhs-note/.
Copy config.example.yaml to {runtimeRoot}/config.yaml.
Add image provider keys only if automatic image generation is needed.
Run python3 scripts/validate_skill.py after setup or structural edits.

Skill Directory

Path	Purpose
`references/pipeline.md`	End-to-end workflow and routing
`references/frameworks.md`	Primary framework definitions
`references/drafting-skeletons.md`	Framework-specific draft body shapes
`references/writing-guide.md`	Writing quality rules
`references/title-rules.md`	Title, hook, tag, CTA, and first-screen copy rules
`references/xhs-constraints.md`	Platform constraints
`references/visual-prompts.md`	Visual planning rules
`references/style-selection.md`	Preset recommendation matrix
`references/style-presets.md`	Primary preset system and override mapping
`references/workflows/prompt-assembly.md`	Series prompt assembly contract
`references/cli-reference.md`	Command reference
`references/operations.md`	Setup, onboarding, export, and maintenance notes
`references/style-template.md`	Client configuration template
`references/skill-maintenance.md`	Documentation and validation rules
`{runtimeRoot}/clients/{client}/style.yaml`	Client profile and visual defaults
`{runtimeRoot}/clients/{client}/history.yaml`	Export history
`{runtimeRoot}/clients/{client}/playbook.md`	Optional local writing notes
`toolkit/dist/*.js`	Built CLI entrypoints
`scripts/*.py`	Intake and validation helpers

runtimeRoot resolves in this order:

./.ls-xhs-note/
~/.liusir-skills/ls-xhs-note/
the skill install directory as a read-only legacy fallback

Pipeline Overview

Step	What happens
1	Load client config and route the request
2	Gather or accept a topic angle
3	Choose the angle and framework
4	Draft the note and polish posting copy
5	Generate images
6	Keep a placeholder post-generation step for later adjustment

Workflow Layers

The skill currently has two layers:

main workflow steps
standalone maintenance tasks

Main workflow steps:

load client configuration
topic intake
angle selection and framework choice
draft the note and polish posting copy
generate images
keep a placeholder post-generation step for later adjustment

Standalone maintenance tasks:

client onboarding
history maintenance

Supported entry modes:

topic-driven drafting
existing draft polishing and export
explicit --step routing

Critical Quality Rules

Read references/writing-guide.md before drafting.
Keep one dominant angle per note.
Write native note copy with model reasoning, not with programmatic skeleton generation as the primary path.
Respect {runtimeRoot}/clients/{client}/style.yaml tone and blacklist fields.
Keep the first screen fast to scan.
Use references/frameworks.md, references/drafting-skeletons.md, references/writing-guide.md, and references/title-rules.md together in Step 4.
Treat Step 5 as an image-generation workflow, not a dedicated first-page shortcut.
If image generation fails, keep prompt-only output and continue.
Use preset as the primary visual decision; open style and layout only as overrides.
Keep command behavior aligned with references/cli-reference.md.
Do not move title, hashtag, slicing, or prompt decisions into heuristic CLI code.

Resilience: Never Stop on a Single-Step Failure

If topic signals fail, continue with a direct topic.
If client configuration is missing, initialize the minimum client structure and continue.
If image generation fails, keep prompt files and generation diagnostics.
Stop entirely only when required input is missing and no safe fallback exists.

Operations

Use references/operations.md for:

first-run setup
client onboarding
export support
maintenance support

Gotchas

A note that preserves every branch of a long-form source will lose speed and clarity.
Title quality depends on the actual angle, not on louder adjectives.
Dense visual layouts fail quickly when the body copy is not already compact.
The workflow drifts quickly if title, hashtag, slicing, or prompt choices move into heuristic code.

Comparison

Workflow	Best use	Limitation
Full note workflow	Repeatable drafting and image generation	Later post-generation handling is still a placeholder
Raw prompt only	Quick idea exploration	No stable history or image plan
Manual packaging	Maximum manual control	Slow and inconsistent

References

Main workflow: references/pipeline.md
Commands: references/cli-reference.md
Writing: references/writing-guide.md
Visual planning: references/visual-prompts.md
Maintenance: references/skill-maintenance.md