The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow where the agent generates source code based on user-provided goals and requirements, creating a vulnerability surface.
Ingestion points: Instructions in SKILL.md guide the agent to 'Clarify goals, constraints, and required inputs' from the user to drive the project structure.
Boundary markers: The skill lacks explicit delimiters or instructions to help the agent distinguish between its own logic and potentially malicious instructions embedded in user-provided goals.
Capability inventory: The agent is tasked with code generation and file system operations, which are sensitive capabilities if influenced by adversarial input.
Sanitization: There is no mechanism described for sanitizing or validating user input before it is incorporated into the generated application logic.
[METADATA_POISONING]: The skill uses the term 'production-ready' in its name and description, which is misleading given the insecure defaults provided in the boilerplate samples.
Evidence: The CORS middleware example in resources/implementation-playbook.md is configured with allow_origins=["*"]. This setting allows any domain to access the API, which is a security risk that contradicts the claim of being ready for production use.

fastapi-templates