The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes various external data sources which could be used as vectors for indirect prompt injection.
Ingestion points: The skill instructions (SKILL.md) explicitly direct the agent to process "published articles", "notes or internal memos", "transcripts", "screenshots", and "prior posts" as the primary source material for content generation.
Boundary markers: There are no explicit instructions to use delimiters, tags, or specific markers to separate untrusted source material from the agent's core instructions, nor are there warnings to ignore embedded commands within that material.
Capability inventory: The skill possesses text generation and repurposing capabilities. It also references the x-api skill for "publishing approved X output", creating a functional chain where potentially influenced content could be staged for publication.
Sanitization: The skill does not define any input validation, sanitization, or filtering mechanisms for the ingested source data, relying instead on a "Quality Gate" that checks for tone and claims rather than instruction injection.

content-engine