Why do the secrets scattered across my project end up inside my coding agent's prompt?

AI coding agents read entire project directories before generating or editing code. That context routinely includes .env files, config YAML, shell history, and CI secrets, all of which get packed into the prompt sent to an external LLM API. On the code-generation side, agents write code that hardcodes the credentials they saw in context. GitGuardian's 2026 State of Secrets Sprawl report found 28.65 million new hardcoded secrets pushed to public GitHub in 2025, up 34% year over year, with AI-assisted commits leaking secrets at roughly twice the rate of human-written commits, and 24,008 unique secrets found sitting in public MCP configuration files. No coding agent today ships a secrets-aware context filter that strips or vaults sensitive values before they enter the prompt.