The Agent Skills Directory

[PROMPT_INJECTION]: The skill extracts and processes data from external URLs (Instagram, TikTok, YouTube, etc.) which serves as a vector for indirect prompt injection.
Ingestion points: Data enters via the source_post.url and subsequent extraction of transcripts, captions, and on-image text.
Boundary markers: The workflow lacks delimiters or instructions to treat ingested content as untrusted data, increasing the risk that embedded malicious instructions could influence the agent or sub-skills.
Capability inventory: The skill possesses the ability to write to the file system in the output/ directory and orchestrate other specialized skills.
Sanitization: No validation or sanitization of the extracted external text is performed before it is passed to the marketing-expert or humanizer skills.
[COMMAND_EXECUTION]: The orchestration includes an image-creator tool that performs HTML-to-PNG rendering. If the HTML structure incorporates unvalidated text from the source posts, it could potentially expose the rendering engine to server-side vulnerabilities like SSRF or unauthorized file access.

creative-recycle