Vibe Coding Discover

AI Coding

token-goat

View on GitHub

Token burn reducer and focus keeper for Claude Code, Codex, Copilot, Gemini CLI, and more: surgical read hints, PDF/Office/CSV/markdown file interception, 205+ filter & interception rules, compact manifest injection, image shrinking, cache and compact skills, cache MCP calls, prompt injection protections, and much more.

★ 12618 forksTypeScriptCustomDFKHelper

A TypeScript CLI and hook layer for AI coding agents that reduces context usage by intercepting repeated reads, filtering command output, compacting images and documents, and preserving session context. Supports multiple coding-agent CLIs and includes prompt-injection protections.

Use Cases

Reduce token usage and cost in AI coding sessionsPrevent agents from repeatedly reading unchanged filesExtract relevant symbols or file sections instead of entire filesFilter build and test output to surface errorsCompress screenshots before sending them to an agentPreserve useful session context across compactionCache repeated skill and web-fetch contentGuard against prompt injection in fetched or extracted content

Built With

Language
TypeScript
Frameworks
Claude Code · Codex CLI · Gemini CLI · Copilot CLI · Aider · Cursor · Cline · Windsurf · OpenCode · MCP

Tags

token-optimization · context-window · AI coding agents · CLI · hooks · prompt-injection protection · file interception · output filtering · image compression · session compaction · skills caching · MCP · cross-platform