What problem does it solve?
Long-running AI tasks often lose context as multiple tools and sessions are used. This skill anchors the objective in a lightweight goal.md and stores persistent findings in findings.md to prevent drift and preserve important reasoning.
Core Features & Use Cases
- Goal anchoring: Injects the current objective into the attention window before tool usage, ensuring continuity across steps.
- External memory: Persists research findings and decision rationale in findings.md for cross-session reference.
- 2-Action rule: Enforces updates after two relevant actions to avoid memory gaps, especially after browsing or viewing multimodal content.
- Pre-tool refresh: Uses PreToolUse hooks to refresh the target before each operation.
- TodoWrite collaboration: Works with the platform's TodoWrite to separate progress tracking from memory.
- Use case: For multi-step projects that span days, such as planning a complex feature and recording decisions.
Quick Start
Create a goal.md with a concise objective and constraints in your project, then add findings.md to persist research and let the hook inject the goal before each tool use.
