The Agent Skills Directory

[SAFE]: No security issues were detected. The skill provides extensive documentation for secure operation, including privacy-focused local modes and clear boundaries for data handling.
[EXTERNAL_DOWNLOADS]: The skill utilizes the @agent-native/core Node.js package and references the official hosted service at plan.agent-native.com for planning functionality.
[COMMAND_EXECUTION]: The instructions include usage of CLI commands via npx @agent-native/core for setup, authentication, and running local plan verification and serve operations.
[DATA_EXFILTRATION]: Plan content is sent to the hosted service at plan.agent-native.com as part of the tool's intended document-sharing functionality. A 'Local-Files Privacy Mode' is explicitly documented to allow users to prevent external data transfer.
[PROMPT_INJECTION]: The skill manages a surface for indirect prompt injection by processing external text plans.
Ingestion points: The planText argument in the create-visual-plan tool (SKILL.md).
Boundary markers: Instructions mandate that published plans be rewritten as standalone proposals, effectively isolating them from conversation history (SKILL.md).
Capability inventory: Interaction with hosted MCP tools for plan storage and local bridge commands for file management (SKILL.md, local-files.md).
Sanitization: HTML wireframes and diagrams are re-sanitized by the platform renderer during updates (wireframe.md).

visual-plan