architect
Ask when consequential design choices must be settled before or during implementation — investigates context, compares…
The genome is Sprout's git-backed runtime state. It stores agent overlays, workspace tools, routing rules, long-term memories, prompts, metrics, logs, and derived cache metadata under a user data directory.
How it fires
How this agent gets triggered: by you, by Claude, or both.
Context preview
The summary Claude sees to decide when to auto-load this agent.
The genome is Sprout's git-backed runtime state. It stores agent overlays, workspace tools, routing rules, long-term memories, prompts, metrics, logs, and derived cache metadata under a user data directory.
The genome is Sprout's git-backed runtime state. It stores agent overlays, workspace tools, routing rules, long-term memories, prompts, metrics, logs, and derived cache metadata under a user data directory.
Typical directories:
Source of truth:
Root files are immutable product defaults. Genome files are mutable runtime overrides. Agent lookup checks the genome overlay first, then root. Workspace tools follow the same pattern: genome tool wins, root tool is fallback.
Source of truth:
Genome mutations are committed to a local git repo. Agent changes, routing rules, prompt updates, memory writes, and maintenance operations produce commits. Rollback uses git revert, not file snapshots.
Source of truth:
Memory JSONL is source of truth. Embeddings and SQLite/FTS indexes are derived and rebuilt after writes. Active recall excludes archived and superseded memories. Explicit compaction can remove inactive rows after git has the audit history.
Source of truth:
Routing rules are not memories. They are decision metadata returned from recall as `routing_hints` and rendered into the prompt. The learn pipeline can create rules through the existing `RoutingRule` mutation path.
Source of truth:
Memory writes use a directory lock under `.cache` and restore snapshots when save, index rebuild, or commit fails. This protects JSONL source files and the derived memory index from partial writes.
Source of truth:
A self-improving, multi-agent AI system that decomposes tasks and delegates to specialist agents. Sprout is an autonomous coding assistant built on a recursive agent architecture — where every action is a goal-directed delegation to a subagent.
Ask when consequential design choices must be settled before or during implementation — investigates context, compares…
Memory investigation, synthesis, and authorized curation when surfaced memories are insufficient
Systematically diagnose and fix bugs: find root cause before attempting any fix
Observe Sprout's live session behavior and send concise guidance when it is drifting, stuck, or missing an important instruction
Explore and understand unfamiliar codebases, map architecture, and synthesize structural insights
Ask when you need to know what Sprout agents, tools, or MCP servers are available, need a reusable capability plan, or need a new…