The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions require the agent to ingest and analyze external data sources, including existing API architectures, database schemas, and backend service patterns. This creates a surface for indirect prompt injection where malicious instructions could be embedded within the files or documentation being reviewed.
Ingestion points: File SKILL.md identifies tasks such as querying context managers for architectures and reviewing current backend patterns.
Boundary markers: There are no explicit instructions or delimiters provided to help the agent distinguish between analyzed data and its own system instructions.
Capability inventory: The persona is granted capabilities to implement code, design APIs, and optimize database schemas based on its analysis.
Sanitization: No sanitization, validation, or filtering mechanisms for the analyzed external content are mentioned.

backend-developer