The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements a session resumption feature that reads content from local markdown files and injects it into the agent's context, creating an indirect prompt injection surface. Ingestion points: Markdown files located in the .omo/handoff/ directory. Boundary markers: Resumed content is wrapped in <RESUME FROM PHASE ...> tags. Capability inventory: The skill uses Read, Bash, Glob, Grep, and dispatches sub-agents via the task() tool. Sanitization: No explicit sanitization or validation of the handoff file content is mentioned before it is interpolated into the session context.
[SAFE]: Access to the local file system is confined to the skill's own references and project-specific coordination folders. No hardcoded credentials or access to sensitive system directories (like .ssh or .aws) were observed.
[SAFE]: No network activity, external code downloads, or remote code execution patterns were identified. All external references are internal to the vendor's own skill ecosystem.

using-meisijiya-skills