Chanjing Ai Creation

Chanjing AI creation Open API client: submit image/video tasks, poll task status, list and get tasks, and optionally download outputs when the user runs down...

Security Scan

The skill is internally consistent: it implements a client for Chanjing's AI creation API, and its scripts, manifest, and instructions align with that purpose; the main security notes are that it reads/writes a local credentials.json and persists access tokens to disk and will only download outputs when explicitly asked.

Purpose & Capability

Name and description match the shipped scripts and manifest. The scripts submit tasks, poll, list/get tasks and optionally download outputs from open-api.chanjing.cc, which is coherent with the stated purpose. No unrelated credentials or services are requested.

Instruction Scope

SKILL.md and scripts explicitly require reading/writing a local credentials.json and calling https://open-api.chanjing.cc, which is expected. The auth helper may open the browser or attempt to invoke a local 'chanjing-credentials-guard' script (via subprocess.run) if present; this is documented in SKILL.md but means the skill can execute a local helper script if the workspace contains it. Downloading results is only performed by download_result.py and is documented as an explicit user action.

Install Mechanism

There is no install spec and the skill is instruction + small Python scripts only. Nothing is downloaded or written to disk by an installer step beyond the scripts already included in the skill bundle.

Credentials

No unrelated environment variables or external credentials are requested. The manifest documents optional CHANJING_OPENAPI_CREDENTIALS_DIR and CHANJING_OPENAPI_BASE_URL. The skill persists access_token and expire_in to credentials.json by design (manifest.persistAccessTokenOnDisk: true); this is proportionate to a CLI API client but increases token exposure risk on multi-user or backed-up hosts.

Persistence & Privilege

alwaysSkill is false and the skill does not request global agent modifications. It reads/writes only its declared credentials file and may open a browser for auth; these behaviors are declared in manifest and SKILL.md.

Assessment

This skill is a normal Open API client for Chanjing. Before installing, note: (1) it will read and write ~/.chanjing/credentials.json (or the directory you set via CHANJING_OPENAPI_CREDENTIALS_DIR) and persist access_token and related fields to disk — avoid using it on shared machines or where that file is backed up to untrusted storage; (2) the auth helper may open your browser or, if present in the workspace, run a local 'chanjing-credentials-guard' script — verify any local helper you trust; (3) downloads only occur when you explicitly run download_result.py, but that script can fetch arbitrary URLs you provide (ensure you trust the URL and that agent network policies are acceptable); (4) network traffic goes to open-api.chanjing.cc (and login pages at www.chanjing.cc) per the manifest. If you need higher assurance, inspect the credentials.json content and the included scripts (they are small and readable) and keep credentials.json permissions restricted (e.g., chmod 600).

功能说明

调用蝉镜 AI 创作 Open API：提交任务、轮询状态、列表与单条查询；仅在用户明确要求时用 download_result.py 可选下载输出。需读写本地 credentials.json 并访问 Open API。跨模型文生图/视频等由 submit_task.py 参数与透传体决定。本 skill 脚本不依赖 ffmpeg/ffprobe。

运行依赖

• python3 与同仓库 scripts/*.py
• 无 ffmpeg/ffprobe 门控

环境变量与机器可读声明

• 环境变量键名与说明：manifest.yaml（environment 段）及本文
• 变量、凭据、合规 permissions、clientPermissions、agentPolicy：manifest.yaml

使用命令

• ClawHub（slug 以注册表为准）：clawhub run chanjing-ai-creation
• 本仓库示例：python skills/chanjing-ai-creation/scripts/submit_task.py …（见正文 Standard Workflow）

---

登记与审稿（单一事实来源）

主凭据、可选 env、下载信任边界等：以 manifest.yaml 为准。本篇从 When to Use 起写业务能力。

When to Use This Skill

当用户要做这些事时使用本 Skill：

• 提交图片或视频 AI 创作任务（submit_task.py）
• 列表 / 单条查询任务（list_tasks.py、get_task.py）
• 轮询异步结果直至完成（poll_task.py）
• 仅在用户明确要求时下载输出（download_result.py）

上述流程依赖读写本地 credentials.json 并调用 open-api.chanjing.cc。

如果需求更接近“文生数字人”，优先使用 chanjing-text-to-digital-person。
如果需求更接近“已有数字人视频合成”，优先使用 chanjing-video-compose。

Preconditions

执行本 Skill 前，必须先通过 chanjing-credentials-guard 完成 AK/SK 与 Token 校验。

本 Skill 与 guard 共用：

• ~/.chanjing/credentials.json
• https://open-api.chanjing.cc

无凭证时，脚本会自动打开蝉镜登录页，并提示配置命令。

审阅与安全（凭据）

与 Purpose / Credentials / Persistence、download_result.py 信任边界相关的逐项说明见 manifest.yaml。

运行时范围：本 Skill 的说明与脚本仅服务于已声明用途——读取本地 credentials.json、调用蝉镜 Open API 端点；不在轮询或查询成功时自动落盘生成物，仅在用户显式执行 download_result.py 时下载资源。

_auth.py 与仓库布局：鉴权辅助可能打开系统浏览器，或按相对路径调用同仓库下的 chanjing-credentials-guard 脚本（例如 open_login_page.py）。这假设当前工作区为 chan-skills 式多 Skill 目录结构，或已单独安装并具备等效路径的 guard skill；否则将回退为仅打开登录页 URL。

敏感数据与持久化：预期仅处理 credentials.json 中的 app_id、secret_key、access_token（及与 token 生命周期相关的字段，见 manifest.yaml）。将 access_token 写入磁盘 是有意设计且已文档化，对 API 客户端而言通常可接受；但在多人共用主机、全盘备份/同步到不可信存储等场景下，磁盘上的 token 会扩大泄露面，需按环境自行评估。

Standard Workflow

AI 创作的主接口是统一提交器：

1. 调用 submit_task.py 提交图片或视频生成任务，得到 unique_id
2. 调用 poll_task.py 轮询直到成功，得到 output_url
3. 如需回看任务参数或错误原因，调用 get_task.py
4. 如需看历史记录，调用 list_tasks.py
5. 只有在用户明确要求保存到本地时，才调用 download_result.py

这个 skill 默认做成“通用任务提交器”：

• 对常见图片/视频模型，优先使用脚本提供的通用参数
• 对特殊模型参数，使用 --body-file 或 --body-json 透传完整请求体

Covered APIs

本 Skill 当前覆盖：

• POST /open/v1/ai_creation/task/submit
• POST /open/v1/ai_creation/task/page
• GET /open/v1/ai_creation/task

Scripts

脚本目录：

• skills/chanjing-ai-creation/scripts/

脚本	说明
_auth.py	读写 credentials.json、获取或刷新 access_token
submit_task.py	提交 AI 创作任务，输出 unique_id
get_task.py	获取单个任务详情
list_tasks.py	列出图片或视频任务
poll_task.py	轮询任务直到完成，默认输出第一个结果地址
download_result.py	下载图片或视频到 outputs/ai-creation/

Usage Examples

示例 1：Seedream 3.0 文生图

TASK_ID=$(python3 skills/chanjing-ai-creation/scripts/submit_task.py \
  --creation-type 3 \
  --model-code "doubao-seedream-3.0-t2i" \
  --prompt "赛博朋克城市夜景，霓虹灯，雨夜，电影镜头" \
  --aspect-ratio "16:9" \
  --clarity 2048 \
  --number-of-images 1)

python3 skills/chanjing-ai-creation/scripts/poll_task.py --unique-id "$TASK_ID"

示例 2：腾讯 Kling v2.1 Master 图生视频

TASK_ID=$(python3 skills/chanjing-ai-creation/scripts/submit_task.py \
  --creation-type 4 \
  --model-code "tx_kling-v2-1-master" \
  --ref-img-url "https://res.chanjing.cc/chanjing/res/aigc_creation/photo/start.jpg" \
  --ref-img-url "https://res.chanjing.cc/chanjing/res/aigc_creation/photo/end.jpg" \
  --prompt "角色从静止到转身，镜头平滑移动，叙事感强" \
  --aspect-ratio "9:16" \
  --clarity 1080 \
  --quality-mode pro \
  --video-duration 5)

python3 skills/chanjing-ai-creation/scripts/poll_task.py --unique-id "$TASK_ID"

示例 3：直接透传完整 JSON

python3 skills/chanjing-ai-creation/scripts/submit_task.py \
  --body-file ./payload.json

Download Rule

下载是显式动作，不是默认动作：

• poll_task.py 成功后应先返回远端 output_url
• 不要自动下载结果文件
• 只有当用户明确表达“下载到本地”“保存到 outputs”“帮我落盘”时，才执行 download_result.py

Output Convention

默认本地输出目录：

• outputs/ai-creation/

Additional Resources

更多接口细节见：

• skills/chanjing-ai-creation/reference.md
• skills/chanjing-ai-creation/examples.md