Holocron

Holocron — diff, lint, snippets, find/replace, history, AI agents and Git for Blip Builder devs.

## Detailed description Holocron is a productivity toolkit for chatbot developers using Blip Builder. It adds the things a flow editor is missing when your bot stops being small: a way to see what you changed before publishing, a way to find what is broken before your customer does, and a way to find where a variable actually comes from. 🔍 SEMANTIC DIFF before publishing — see exactly what changed (blocks, conditions, variables, scripts) since your last snapshot, instead of publishing blind. 🩺 FLOW LINT — over 35 checks across structure, variables, integrations, scripts and AI agents: orphan blocks, unreachable states, undefined variables, broken agent handoffs, self-loops, HTTP instead of HTTPS, hardcoded secrets and URLs, unguarded JSON.parse, tracking without a category, and more. Problems are marked on the canvas itself, on the block that has them. 🎯 PUBLISH READINESS — your flow gets a score from A to E, broken down by structure, variables, integrations and instrumentation, with technical debt estimated in minutes and whether it improved or got worse since last time. 🔎 FIND ALL REFERENCES — where a variable is written and where it is read, grouped by block, with the exact text around each hit. Tells you when a variable is read but never written, which renders empty in production without the Builder ever warning you. Also works for blocks: who actually reaches this one. 🤖 AI AGENT AUDIT — inspect prompt, handoffs, model config, tools and usage of every agent block, plus a visual graph of how they connect. Flags tools with no description or an invalid schema, models the platform has deprecated, temperature that is too high for an agent that only routes, and Prompt Shield left off. 📊 INSIGHTS — 26 analyses across 7 lenses: structure, conversation, data and variables, integrations, WhatsApp, AI agents and readiness. Token cost estimates, prompt review, dead and ghost variables, dead ends, dependency maps. 🔁 FIND & REPLACE across messages, conditions, script source and titles, with regex and case-sensitive options, and a preview before anything is written. ⬆️ BULK OUTPUTS — create the same conditional output across many blocks at once, with a preview that tells you how many blocks change and how many already had it. 📦 SNIPPETS — save any block as a template and reuse it across bots. Copy-paste compatible with native Blip Builder. 🕒 VERSION HISTORY — automatic snapshot on every publish, plus manual saves. Compare any two versions side by side. ⬆️ GIT INTEGRATION — auto-commit flow.json to GitHub, GitLab, Azure DevOps or any custom webhook on every publish. 🗺 CANVAS TOOLS — minimap, block tree explorer, focus mode to isolate one block's neighbourhood, block comparison field by field, and Building blocks: colored areas that group the canvas, with custom categories and an owner, URL and description per area. 📥 IMPORT / EXPORT — move flows between bots (HMG → PRD) with flow, variables, tags, Building blocks and extension settings. Export to Markdown or Mermaid. 📊 STATUS BAR — a persistent footer showing block count, invalid blocks, snapshot age and Git status, with one-click access to every tool. ⌨️ KEYBOARD SHORTCUTS — Ctrl+Shift+P opens the command palette. Ctrl+K jumps to any block. Ctrl+Shift+L lint, F find & replace, U references, J history, M tools menu, G go to block, ? shortcut list. Interface available in Portuguese, English and Spanish. No data is collected by this extension. Git tokens are stored locally in your browser only, and nothing is sent anywhere except the Git provider you configure yourself. See the privacy policy for details. ## (fim do bloco) --- ## Notas de versão (campo separado da loja) Sugestão para o campo de release notes, curta de propósito: New: Find All References for variables and blocks — see where a variable is written, where it is read, and what is only a guess. Fixed: two false positives that hit every bot using AI Agents. A Builder default variable such as {{agent.fullName}} was reported as undefined, and an agent's output variable was not recognised as a write. Ghost variable counts drop accordingly. Fixed: reading a variable without writing it is no longer reported as broken when the bot uses subflows — the value can legitimately come from the parent flow.