功能描述

蝉镜内容创作聚合技能包。提供凭据管理、TTS 语音合成、声音克隆、数字人口播、对口型、文生图/视频、定制数字人训练、一键成片编排、卡通视频编排等能力。当用户表达"做一个短视频""语音合成""数字人口播""一键成片""卡通视频"等意图时触发。副作用：HTTPS 访问蝉镜 Open API、读写本地凭据文件、下载、f...

使用说明 (SKILL.md)

chanjing content creation skill

Name: Chanjing Content Creation Skill
Author: binkes

蝉镜内容创作技能包（顶层路由）

本文件只负责：触发边界、路由消歧、统一执行约束与运行时契约摘要。
具体执行步骤以子技能 *-SKILL.md 为真值；长文细节以 references/ 为真值。

Use When

当用户需求涉及以下任一项时使用本 Skill：

短视频成片（含一键成片、卡通视频、口播混剪）
蝉镜语音能力（TTS、声音克隆）
数字人能力（公共/定制数字人、对口型、文生数字人）
蝉镜 AI 创作（文生图、文生视频）
蝉镜凭据管理（APP_ID/SECRET_KEY 配置、Token 刷新、登录引导）

Do NOT Use

与蝉镜平台无关的通用需求（纯文案、非蝉镜剪辑工具等）
用户仅讨论方案且明确不执行任何蝉镜 API/脚本调用

Route Map（唯一映射）

用户意图	路由目标
一键成片、完整短视频、做一个 xx 视频	`orchestration/chanjing-one-click-video-creation/chanjing-one-click-video-creation-SKILL.md`
卡通短剧、动漫多镜叙事	`orchestration/cartoon-video-creation/cartoon-video-creation-SKILL.md`
文字转语音、语音合成	`products/chanjing-tts/chanjing-tts-SKILL.md`
声音克隆、音色复刻	`products/chanjing-tts-voice-clone/chanjing-tts-voice-clone-SKILL.md`
对口型、唇形驱动（非卡通）	`products/chanjing-avatar/chanjing-avatar-SKILL.md`
数字人口播、公共数字人、纯口播视频、卡通数字人口播合成	`products/chanjing-video-compose/chanjing-video-compose-SKILL.md`
文生图、创意视频	`products/chanjing-ai-creation/chanjing-ai-creation-SKILL.md`
定制数字人（上传素材训练）	`products/chanjing-customised-person/chanjing-customised-person-SKILL.md`
人设图、人像 LoRA、文生数字人形象	`products/chanjing-text-to-digital-person/chanjing-text-to-digital-person-SKILL.md`
凭据、APP_ID/SECRET_KEY、Token	`products/chanjing-credentials-guard/chanjing-credentials-guard-SKILL.md`

Conflict Resolution（强制消歧）

出现重叠意图时，按以下优先级判定：

完整成片（文案+分镜+画面+交付）优先路由 chanjing-one-click-video-creation
卡通/动漫多镜叙事成片优先路由 cartoon-video-creation
卡通数字人口播合成（model=2）或仅数字人口播，路由 chanjing-video-compose
已有视频改口型（非卡通）路由 chanjing-avatar
路由不确定时，遵循 orchestration/orchestration-contract.md 的 need_param 分支最小补参

Root Workflow（最小入口流程）

Step	动作	输出	下一步	失败分支
1	凭据可用性检查（缺失时走 `chanjing-credentials-guard`）	`ok` 或 `auth_required`	Step 2	`auth_required`：先完成鉴权再回 Step 2
2	按 Route Map 选择单一目标子技能	目标 `*-SKILL.md` 路径	Step 3	`need_param`：只补路由必需信息
3	打开目标子技能并执行其 Standard Workflow	子技能结果	Step 4	按子技能失败分支执行
4	返回产物 URL/路径或标准错误语义	`ok`/`need_param`/`auth_required`/`upstream_error`/`timeout`	结束	需要降级跨产品时先征得用户确认

Global Execution Rules

仅可调用 products/*/scripts/ 与 orchestration/*/scripts/ 的已有入口。
禁止自行编写临时脚本替代既有 CLI。
默认不主动下载；仅当用户明确要求“下载/保存到本地”时执行下载脚本。
重复任务可复用已确认参数，但当核心输入冲突时必须切换为新任务。
外部副作用（网络、写盘、子进程、浏览器）必须与运行时契约一致。

Three-Layer Architecture

层级	路径	作用
顶层入口	`SKILL.md`	路由、消歧、契约摘要
L1 公共层	`common/`	公共鉴权、HTTP 封装、上传下载、日志
L2 产品层	`products/`	单产品能力与标准脚本
L3 编排层	`orchestration/`	跨产品场景编排

依赖方向固定为：common \x3C- products \x3C- orchestration（禁止反向依赖）。

Runtime Contract（摘要）

完整契约见 references/top-level-runtime-contract.md。

凭据来源：项目 .env（CHANJING_APP_ID、CHANJING_SECRET_KEY）
凭据文件：skills/chanjing-content-creation-skill/.env
Token 策略：不写入磁盘；按需请求并在当前进程内短时复用
API 基址：${CHANJING_API_BASE:-https://open-api.chanjing.cc}
关键环境变量：CHANJING_APP_ID、CHANJING_SECRET_KEY、CHANJING_ACCESS_TOKEN、CHANJING_TOKEN_EXPIRE_IN、CHANJING_API_BASE、CHAN_SKILLS_DIR（其中 token 变量仅作为进程级可选输入，不持久化）
外部二进制：python3（所有脚本入口）、ffmpeg/ffprobe（主要用于 L3 成片链路）
副作用分类：出站 HTTPS、本地写入、子进程、浏览器引导登录

Constraints

脚本写盘边界受运行时契约约束（凭据路径、--output-dir、output/、work/）。

安全使用建议

This skill appears to do what it says: route intents to Chanjing product scripts and orchestrate media creation. Before installing, consider: 1) Provide Chanjing API credentials (CHANJING_APP_ID and CHANJING_SECRET_KEY) only if you trust the Chanjing service; the skill will read them from env or write them to skills/chanjing-content-creation-skill/.env. 2) Outputs and downloads: the skill may download result URLs returned by the API into the skill's output directories and will invoke ffmpeg for composition — ensure you run it in an environment where file writes and subprocesses are acceptable. 3) Network scope: primary network host is open-api.chanjing.cc (and www.chanjing.cc for login); note that API-provided signed URLs may point to other hosts for upload/download (this behavior is documented in SKILL.md and manifest). 4) Credentials hygiene: use a dedicated/trusted API key, avoid sharing production secrets, and be prepared to rotate keys if you stop using the skill. 5) Minor documentation mismatch: manifest lists the Chanjing keys as optional while the SKILL.md treats them as required — treat them as required for real use. If you need higher assurance, run the package in an isolated environment (container or throwaway VM) and/or review the specific product scripts you expect to use (products/* and orchestration/*) before giving it access to sensitive data.

功能分析

Type: OpenClaw Skill Name: chanjing-content-creation-skill Version: 1.0.3 The skill bundle is a legitimate and well-architected integration for the Chanjing (蝉镜) AI content creation platform. It manages API credentials locally in a .env file using restricted file permissions (0o600) and directs all network traffic to documented platform endpoints (open-api.chanjing.cc). The bundle includes proactive security features such as input sanitization and multi-step procedural validation scripts (extract_params.py and validate_step.py) to ensure data integrity. Subprocess usage is strictly limited to internal Python scripts and standard media utilities (ffmpeg/ffprobe), with no evidence of malicious execution, data exfiltration, or harmful prompt injection.

能力标签

requires-oauth-token

能力评估

✓ Purpose & Capability

Name/description (Chanjing content creation: TTS, voice-clone, digital humans, video orchestration) match the codebase and runtime instructions. Declared binaries (python3, ffmpeg/ffprobe) and required env vars (CHANJING_APP_ID, CHANJING_SECRET_KEY) are directly used by the libraries and CLI scripts and are expected for this functionality.

ℹ Instruction Scope

SKILL.md is a top-level router and explicitly delegates operations to product/orchestration sub-skill scripts (products/*, orchestration/*). It documents the side-effects it will perform: HTTPS to Chanjing Open API, reading/writing local credentials (.env under skill dir), downloads of result URLs, launching subprocesses (python/ffmpeg) and optionally opening a browser for login. It also instructs using web_search for trend research in orchestration docs — that is out-of-skill network activity but described as part of content-research steps. Behavior described in SKILL.md aligns with the code; there is no hidden instruction to access unrelated system paths or exfiltrate arbitrary secrets beyond the Chanjing credentials and tokens.

✓ Install Mechanism

No install spec — the skill is instruction + bundled Python scripts. All code is included in the package (L1/L2/L3). No remote downloads or install URLs are executed at installation time, reducing supply-chain risk. Scripts will be executed via python3 subprocesses at runtime.

ℹ Credentials

Requested environment variables (CHANJING_APP_ID, CHANJING_SECRET_KEY) are appropriate for an API-integrating skill. The runtime also uses optional token-related env vars and CHANJING_API_BASE; those are documented in SKILL.md. Minor inconsistency: manifest.yaml lists those variables under optional, while SKILL.md and metadata indicate they are required for operation — you should assume you need to provide app_id/secret. The skill reads/writes skills/chanjing-content-creation-skill/.env (declared in runtime contract).

✓ Persistence & Privilege

always: false and disable-model-invocation defaults are standard. The skill writes its own .env file under its skill directory (documented) and does not claim to modify other skills or global agent settings. It may open a browser to guide login (documented).

版本历史

v1.0.3

- 更新运行时契约：Token 策略改为“不写入磁盘，仅短时复用于当前进程”。 - 明确环境变量：Token 相关变量仅为进程级可选输入，不再要求持久化。 - 补充并细化运行时契约部分描述，强调凭据与 Token 的生命周期与存储边界。 - 保持核心路由逻辑、典型使用场景与全局约束不变。

v1.0.2

- Clarified the top-level routing and capability summary for the chanjing content creation skill. - Added a detailed route map, mapping user intents to specific subskills. - Defined clear boundaries for use, including “Do NOT Use” scenarios. - Outlined conflict resolution priorities and minimum required execution flow. - Specified global execution rules, including credential checks and strict side effect management. - Documented architectural layers and the runtime contract in detail. - Introduced new sections for workflow, dependency direction, and constraints.

v1.0.1

- Adds a new "运行时契约" (runtime contract) section detailing expected behavior and side effects when running scripts, covering credentials management, environment variables, binaries, local file writes, and network I/O. - Clarifies the top-level SKILL.md purpose: now includes both routing, hierarchy explanation, and explicit runtime contract; reiterates that all business logic and parameter details remain in lower-level docs and scripts. - Documents typical environment variables (e.g. CHANJING_API_BASE, CHANJING_CONFIG_DIR, CHAN_SKILLS_DIR), external binary dependencies (ffmpeg, ffprobe), and file write locations for all products and orchestration scripts. - States that the skill does not auto-load by default (always: false); clarifies enabling behavior. - Reinforces separation of responsibilities between SKILL.md, scripts, and product/orchestration layers for safer, more predictable usage. - No changes to APIs, CLI parameters, or scripts themselves—documentation and meta structure only.

v1.0.0

- Initial release of chanjing-content-creation-skill. - Provides a single top-level SKILL.md as the entry point for content creation skills, focusing solely on routing and responsibility description. - No business logic or execution details included at the root; detailed instructions and scripts are organized per product or scenario subdirectory. - Clear directory and dependency structure: common utilities, product modules, and orchestration flows are physically separated with strict dependency rules. - Comprehensive routing table and navigation guidance, mapping user intent and keywords to the correct documentation or script location.

元数据

Slug chanjing-content-creation-skill

版本 1.0.3

许可证 MIT-0

累计安装 1

当前安装数 1

历史版本数 4

常见问题