The Agent Skills Directory

[SAFE]: The skill defines a production-grade framework for backend development foundations, including bootstrap ordering, connection pooling, and graceful shutdown. It relies on high-quality technical documentation from well-known and trusted sources to guide the agent toward secure and deterministic architectural decisions.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves analyzing external project data such as deployment manifests, environment variables, and CI/CD definitions. This data ingestion represents a surface for indirect prompt injection if a target project contains adversarial instructions. The risk is mitigated by the skill's explicit focus on schema-based validation and fail-fast logic to prevent non-deterministic behavior. Evidence: 1. Ingestion points include project manifests and environment variables (SKILL.md workflow step 4 and Paper 001 section 3). 2. Boundary markers are not explicitly defined for data interpolation. 3. Capabilities include reading project files using agent tools. 4. Sanitization of external input is handled via the recommended use of standards-aware parsers.

runtime-delivery