AI coding

Managing context token budgets

• Enumerating What an Agent Sends to a Check
• Token Math for Long Files: Counting a Codebase Before You Load It
• What to Prune When a Context File Outgrows Its Token Budget
• Where a Check's Prompt Grows: Context Drift Across Runs
• Alerts on a Slope, Not a Value: Thresholds for Growing Context
• Caching Token Counts Across Runs: A File Hash Registry
• Cap, Summarize or Drop: A Decision Rule for Each Prompt Segment
• Chunking a Long Context File for a Small Model
• Costing Each Section of a Context File
• Estimating Tokens for a Monorepo: Sampling vs Counting Everything
• Loading Conventions Only for the Files a Task Touches
• Reading a Summarizer's Own Token Bill
• Rewriting a Context Section Without Changing Pass Rate
• Token Cost of Loading a Call Graph into Agent Context
• What to Load First: Context Packing for a Legacy Codebase Agent
• When a Context File Is Loaded: Triggers and Cost
• When a File List Is Too Long: Pruning Before the Preflight
• When to Summarize a File and When to Retrieve a Slice