The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as its primary function is to process user-provided descriptions, schemas, and requirements to generate functional source code.
Ingestion points: User requests such as "create an API endpoint" or "build a REST API" are ingested to determine the "user goal" (SKILL.md).
Boundary markers: The skill does not implement specific delimiters (e.g., XML tags or clear start/stop markers) to isolate user-provided data from instructions, potentially allowing adversarial inputs to influence the agent's behavior.
Capability inventory: The skill is configured to use code generation and CLI tools, including codex-cli and claude-code, giving it the ability to create and modify files (SKILL.md).
Sanitization: While the skill instructs the agent to include validation in the generated code, there are no instructions for the agent to sanitize or filter the user's initial prompt content before processing it.
[SAFE]: The skill instructions and examples demonstrate a focus on secure development practices.
Code examples include proper input validation logic and the use of bcrypt for secure password hashing (SKILL.md).
The centralized error handler example utilizes process.env.NODE_ENV to selectively hide stack traces and sensitive internal error details from end-users in production environments, which is a standard security best practice.
Provenance information, including the source repository and commit hash, is clearly documented in ORIGIN.md and metadata.json for auditability.

api-endpoint-builder