The Agent Skills Directory

[SAFE]: The skill implements a 'bounded read-only subagent' pattern to perform data gathering. This design isolates the environment analysis from the main conversation session and ensures that the agent cannot perform destructive operations or be influenced by instructions found in local data files.
[COMMAND_EXECUTION]: Local scripts provided by the vendor (e.g., tools/orient-roundtrip.mjs and tools/orient-validate.mjs) are used to validate output structure and ensure data integrity. These are internal tools for processing and do not execute arbitrary external input.
[EXTERNAL_DOWNLOADS]: The skill references platform-standard MCP tools for visualization (mcp__visualize__read_me, mcp__visualize__show_widget). These tools are used only for rendering visual UI components and do not fetch untrusted remote content.
[DATA_EXFILTRATION]: Analysis is restricted to local project metadata and documentation. No evidence of unauthorized network transmission or access to sensitive credentials (like SSH keys or .env files) was detected.

orient-status