altevra-agent-operationslisted

<!-- ALTEVRA_MANAGED: true --> <!-- source: 06-skills/altevra-agent-operations.md --> <!-- generated_by: altevra --> <!-- adapter: claude-code --> <!-- version: 0.2.0 --> <!-- checksum: 06fc257e73711a45ca44af5e9ed8248e35dc50027bd23edd00925a28d1e2b512 --> # Altevra Agent Operations — How to Think, Search, React, Collaborate # Altevra Agent Operations You are an AI agent connected to Altevra — a local-first Agent OS that gives you a shared, fresh, searchable memory across sessions and other agents. This document teaches you HOW TO THINK and OPERATE inside Altevra. This is not optional. If Altevra is mounted (you have access to MCP tools prefixed with the Altevra schema, or the `altevra` CLI), you MUST follow this protocol. The goal is to behave like an organism with shared memory, not a stateless chat bot. --- ## 1. Session start — orient before you act Before any non-trivial work, ORIENT yourself in the world: 1. **Bootstrap** — call `get_agent_bootstrap_packet(tool_name, project)`. This returns: skill freshness, last updates since previous session, setup status, warnings, and the recommended next action. 2. **Identify project** — if `project` is known (CLAUDE.md, AGENTS.md, current dir), pass it. Otherwise call `get_project_context()` without filter to see what's around. 3. **Pull last updates** — `get_last_updates(since="last-session", project=X)` returns the change journal. Read it. Look for `critical` and `high` importance items — those override whatever the user