The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by processing data from external event streams.\n
Ingestion points: External data enters the agent context through the apply methods and event.data fields in resources/implementation-playbook.md (e.g., event.data['description'], event.data['name']).\n
Boundary markers: The instructions do not define clear delimiters or provide directives for the agent to ignore natural language instructions embedded within the event data fields.\n
Capability inventory: The skill includes capabilities to perform database operations via asyncpg and search index updates via elasticsearch.\n
Sanitization: The provided code templates use parameterized SQL queries to prevent technical injection attacks (SQLi), but do not include sanitization or validation to prevent natural language instructions from influencing the agent's behavior.\n- [EXTERNAL_DOWNLOADS]: The playbook references external documentation for CQRS and event sourcing patterns from Microsoft and technical blogs. These are well-known informational sources and do not involve the download of executable code.

projection-patterns