The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill handles external data that could potentially contain malicious instructions if the source (Kling AI task results) is compromised.
Ingestion points: The video_url and prompt parameters in the download_video and generate_and_store functions (SKILL.md, references/aws-s3-integration.md).
Boundary markers: No specific delimiters or warnings are used to separate untrusted data from the agent's instructions.
Capability inventory: The skill uses file write operations (open), network requests (requests.get), and cloud storage SDKs (boto3, google-cloud-storage, azure-storage-blob) to move data (SKILL.md).
Sanitization: No explicit sanitization of the input URLs or metadata is implemented beyond standard SDK error handling.
[EXTERNAL_DOWNLOADS]: The skill contains logic to download binary video files from remote URLs.
The download_video function in SKILL.md uses requests.get to fetch data from a provided video_url and writes it to the local filesystem.

klingai-storage-integration