The Agent Skills Directory

[SAFE]: Comprehensive analysis of the orchestrator's logic and sub-agent instructions confirms that all operations are aligned with its stated purpose. The skill follows best practices for multi-agent coordination.
[COMMAND_EXECUTION]: The skill uses local scripts (state.js, progress.js, check-env.js) and shell commands (git, npm, npx) to manage the development lifecycle. These executions are restricted to project-specific tasks like branching, testing, and linting.
[EXTERNAL_DOWNLOADS]: The skill references several Model Context Protocol (MCP) servers for extended functionality (Linear, GitHub, Web Search). These are downloaded and executed via npx from well-known and trusted organizations.
[PROMPT_INJECTION]: The skill employs specific instructions and iterative loops to ensure sub-agents adhere to the orchestration plan. While it uses strong instructional language (e.g., "IMPORTANT", "CRITICAL"), these are used for internal control flow and do not represent malicious bypass attempts.

buddy