The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting and acting upon untrusted data to generate or modify executable code.
Ingestion points: The workflow (SKILL.md) reads a 'requirements brief' from the conversation context or the requirements-gathering skill. It also fetches existing code from a remote branch using dxs function get.
Boundary markers: No explicit delimiters or instructions are used to distinguish between the agent's instructions and potentially malicious instructions embedded in the requirement briefs or existing code.
Capability inventory: The skill possesses significant capabilities, including local file writing, code validation, and the ability to upload generated TypeScript code to a remote branch via dxs configuration upsert flow.
Sanitization: There is no evidence of sanitization or safety checks performed on the requirements or existing code before they are used to influence the generation of new code logic.

function-creator