The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external data from the Replo platform, including site content, products, orders, and competitor ad research. This creates a potential surface for indirect prompt injection where malicious instructions embedded in external data could influence the agent's behavior.
Ingestion points: Data enters the context via the Replo connector from sources like website pages, product descriptions, and ad research.
Boundary markers: No specific delimiters or instructions to ignore embedded commands are present in the skill definition.
Capability inventory: The skill allows the agent to call tools and load additional instruction sets based on the data it processes.
Sanitization: No evidence of sanitization or validation of the external content is provided in the instructions.
[DYNAMIC_EXECUTION]: The skill uses a dynamic loading mechanism to fetch and execute further instructions.
Evidence: The instructions direct the agent to call the skills tool with a specific name to load full SKILL.md files for sub-tasks at runtime.

replo