Alignment Checkpoint
Establishes a mandatory intent-alignment gate to clarify goals and secure explicit approval before executing high-stakes or ambiguous tasks.
Alignment Checkpoint acts as a safety and precision layer for Claude Code, preventing the AI from running tools or modifying files prematurely. When triggered by ambiguous or high-risk requests, it extracts explicit goals, surfaces hidden assumptions, and defines success criteria in a structured format. The skill presents three distinct implementation paths—minimal, balanced, and comprehensive—and halts all execution until the user provides a specific approval command. This ensures the AI's actions are perfectly synced with user intent, reducing errors in complex engineering workflows.
주요 기능
01Automated goal and assumption extraction in structured JSON.
02Hard execution gate that prevents tool use until explicit user approval.
03Stasis mode for verbatim request preservation and plan-only outputs.
04Single-threaded clarifying questions to prevent communication overload.
05Three-tiered implementation options: minimal, balanced, and comprehensive.
062 GitHub stars
사용 사례
01Clarifying complex or multi-part engineering requests before any code is changed.
02Generating architectural plans without triggering immediate tool executions or file edits.
03Ensuring safety during high-risk operations on production-sensitive codebases.