The Agent Skills Directory

[SAFE]: The skill defines a legitimate design methodology and architectural patterns. It uses domain-specific tools for graph manipulation that are consistent with its stated purpose. Analysis found no evidence of prompt injection, obfuscation, or malicious intent.\n- [INDIRECT_PROMPT_INJECTION]: The skill describes a data ingestion surface for external specifications. Evidence chain: (1) Ingestion point: brainstorming 'spec-files' processed via the 'intake' step. (2) Boundary markers: No explicit prompt delimiters or boundary markers for untrusted data are defined. (3) Capability inventory: Tools for graph modification such as iskron_add_kriya and iskron_arrow. (4) Sanitization: The skill mentions an 'intake' phase, though specific sanitization logic is not detailed within this file. This surface is evaluated as safe as it is confined to domain-specific knowledge management.\n- [EXTERNAL_DOWNLOADS]: The skill references various graph management tools (e.g., iskron_search, iskron_orient). These are internal vendor resources and do not involve connections to untrusted external domains or unauthorized remote code execution.

design