What problem does it solve? Files, audio, images, PDFs, and links arriving through Telegram, email, webhooks, or a web GUI often get dumped straight into a knowledge base, polluting curated memory with uncurated raw material. This Skill enforces a staged intake pipeline so raw input is preserved, extracted, and triaged before anything is promoted into the semantic Acervo. ## Core Features & Use Cases - Unified IntakeEnvelope: Every channel (Telegram, WhatsApp, email, webhook, API, dashboard) converges to a single envelope schema with intake_id, channel, MIME type, and correlation metadata, so channels never dictate workflow. - Five-Phase Standard Flow: Reception into _inbox/{intake_id}/original/, manifest creation with SHA-256 hash, type-specific extraction (transcription, OCR, Markdown conversion, ZIP inventory), cognitive triage with a destination hypothesis, and explicit promotion. - Explicit Inbox Commands: The inbox is only checked on explicit request ("verifique o inbox" or "inseri arquivo X"), and nothing is moved or promoted without user confirmation. - Use Case: You receive a contract PDF from a client by email. The Skill stores the original in _inbox/, extracts the text to Markdown, proposes the correct microverso and contracts/ directory, and waits for your confirmation before promoting it to the semantic Acervo. ## Quick Start Ask the assistant to check the inbox and propose destinations for any new files, for example by saying "verifique o inbox" after dropping a PDF into the intake folder.
