The Agent Skills Directory

[SAFE]: No malicious patterns, suspicious remote connections, or unauthorized command executions were detected. The skill's behavior is consistent with its stated purpose of managing local context for agent invocations.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface due to its core functionality of ingesting external data into the agent's context.
Ingestion points: Accesses files from .cocoplus/grove/patterns/, .cocoplus/context/, .cocoplus/snapshots/, and .cocoplus/prompts/ (SKILL.md).
Boundary markers: Ingested data is structured into a dedicated markdown block with clear headers, although it lacks specific defensive instructions for the model to ignore directives embedded within that data.
Capability inventory: The skill performs filesystem read operations and prepends content to the startup context. No subprocess execution, network access, or sensitive file writes were identified.
Sanitization: The skill ranks and includes file content based on relevance scores but does not perform content sanitization on the files themselves.

cocoscout