The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process potentially untrusted data from various files within the knowledge directory to adapt its output to a specific brand voice and target persona. This creates a surface for indirect prompt injection if those files contain malicious instructions.
Ingestion points: Reads content from knowledge/brand/voice.md, knowledge/icp/personas.md, knowledge/markets/positioning.md, knowledge/services/, and knowledge/content-library/.
Boundary markers: The instructions lack explicit structural delimiters or strict instructions to ignore embedded commands within the ingested text files, relying instead on high-level guidance to mirror rhythm and cadence.
Capability inventory: The skill possesses file system read and write capabilities, specifically writing output files to the output/linkedin-post/ directory.
Sanitization: The skill includes some logical guardrails, such as stop conditions for unauthorized naming of entities and a verification process to ensure metrics are sourced from provided material, which serves as a basic integrity check.

linkedin-post