kdeps

Stream editor that reduces prose to fewer tokens. Pipes text (CLAUDE.md, README, memory, instructions) through the turo binary and injects the reduced output as the system prompt. Deletes filler, collapses definition-like phrases to the word they define, swaps token-cheaper glosses and synonyms, strips stopwords, deduplicates, and lemmatizes — keeping meaning-bearing words. Use when user says "use turo", "compact context", "reduce tokens", or invokes /turo. Auto-triggers when token budget is tight.

Create kdeps components, agents (workflows), and agencies. Use when the user wants to build a kdeps project, write workflow.yaml, component.yaml, or agency.yaml, configure workflow input (api, bot, file), add resources (chat, httpClient, sql, python, exec, scraper, searchWeb, botReply, file, git, codeIntelligence, apiResponse), wire multi-agent pipelines, package/deploy a kdeps app, or publish to kdeps.io.