The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow where the agent interacts with external surfaces such as web browsers, HTTP APIs, and command-line interfaces, creating a surface for potential indirect prompt injection. * Ingestion points: Data from web pages, API responses, and CLI outputs are incorporated into the agent's context during the 'Drive' and 're0-memo' phases. * Boundary markers: The skill does not explicitly define delimiters to separate external data from system instructions. * Capability inventory: The workflow involves generating, executing, and versioning code modules and templates. * Sanitization: No explicit sanitization or instruction-filtering mechanisms are described for data captured from external surfaces.
[SAFE]: The core logic of the skill is purely instructional and focused on iterative development best practices. There is no evidence of obfuscation, credential theft, persistence mechanisms, or unauthorized privilege escalation. The adversarial review mentioned is a quality assurance process for the code produced.

re0-loop