context-guard

What problem does it solve?

The context in AI workflows is a finite resource, and tool outputs consume the majority of the budget. This skill guides token economy at every step by avoiding rereading content already in context, applying targeted reads, and enforcing lean patterns to minimize waste.

Core Features & Use Cases

• Targeted reading over full-file scans (grep) to locate relevant information and avoid unnecessary token usage.
• Safe Bash practices with exit 0 to prevent cascading failures in batch operations.
• Read-once or optional reads with robust fallbacks, parallelization of independent tasks, and externalized write strategies for large results (plans/notes).
• Subagents for heavy explorations with isolated contexts to keep the main conversation lean.

Quick Start

Start by preferring targeted grep reads over full file reads and enable subagents only for heavy explorations.