The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of system commands including curl for API interaction, jq for JSON processing, and ffmpeg/ffprobe for video assembly and quality assurance.
[COMMAND_EXECUTION]: Documentation and setup examples in SKILL.md include sudo commands for configuring local directory permissions and copying files for asset hosting.
[EXTERNAL_DOWNLOADS]: The skill interacts with Volcengine's official API endpoints (ark.cn-beijing.volces.com) to submit video generation tasks and retrieve results.
[PROMPT_INJECTION]: The skill implements an indirect prompt injection surface by reading user-controlled data from Feishu Bitable rows to dynamically construct director prompts for video generation models.
Ingestion points: Data is ingested via lark-cli base +record-get from Feishu Bitable records.
Boundary markers: The skill uses structured templates (e.g., templates/seedance-row24-director-template.md) to wrap content, though explicit "ignore embedded instructions" guards are not detailed for the video model layer.
Capability inventory: The agent has capabilities to execute shell commands (curl, ffmpeg, ffprobe), write to the local filesystem, and update external Feishu records.
Sanitization: The skill employs a 'Dialogue Lock' and a 'Base Asset Ledger Write Guard' script (scripts/base_asset_ledger_audit.py) to verify data integrity and consistency between the Bitable ledger and local manifests.

feishu-seedance-video-pipeline