The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content from multiple local files which could contain adversarial instructions intended to influence the agent's behavior during text generation.
Ingestion points: The skill reads mkdocs.yml, docs/course-description.md, docs/learning-graph/book-metrics.md, and docs/learning-graph/chapter-metrics.md to extract metadata and statistics.
Boundary markers: There are no explicit instructions or delimiters provided to the agent to differentiate between data and potentially malicious commands within the source files.
Capability inventory: The skill utilizes file reading capabilities to populate social media templates and craft announcement variations.
Sanitization: The skill lacks validation or filtering mechanisms for the content extracted from the textbook documentation before it is interpolated into the final prompt context.

linkedin-announcement-generator