The Agent Skills Directory

[SAFE]: The skill serves as a conceptual guide for implementing iterative retrieval loops. Analysis of the markdown and pseudocode confirms no executable scripts, network operations to untrusted domains, or sensitive credential access. The logic presented is purely documentation-based.
[INDIRECT_PROMPT_INJECTION]: The skill describes a workflow that ingests codebase data, which is a potential surface for indirect injection. Ingestion points: Files retrieved based on keywords and patterns in SKILL.md. Boundary markers: None specified in the design pattern. Capability inventory: The logic assumes file read capabilities for context gathering. Sanitization: The pattern focuses on relevance scoring and does not include explicit content sanitization.

iterative-retrieval