多 agent 工程小队配置：Jev 跑机械决策层

damkin @damkina7原推发布于 2026.10.03

在 X 上打开原推

Official Anthropic tip for Claude Code: stop burning Opus 5.5 context on tasks Sonnet 5.5 can swarm, while Fable 5.1 sits idle Initialize a multi-agent engineering pod with /teams claude --teammates frontend,backend,adversary Opus 5.5 functions as the Principal Architect: it defines type boundaries, reviews AST diffs, and controls the merge queue Sonnet 5.5 worker instances spin up isolated git worktrees to churn through business logic and endpoint scaffolding in parallel Fable 5.1 lives on the peer-to-peer event bus as a relentless Zero-Trust Adversary, intercepting code execution at three critical gates: → contract negotiation: does the client payload strictly match the API schema? → test degradation: are we solving the underlying invariant or just suppressing an assertion? → pre-merge review: what race condition, edge case, or injection vector survived the build? Sonnet 5.5 implements. Fable 5.1 audits. Opus 5.5 verifies and deploys Jev engineering runs the identical pattern one layer down: mechanical decisions that need no reasoning, such as which file to open, which tool to invoke, retry or abort, execute in 16ms, so the frontier models only wake up when execution paths actually diverge Architect on high. Execute on medium. Keep your adversary on the p2p bus - the full team setup: > Opus 5.5 on high effort orchestrates the workspace and aggregates pull requests > teammate-frontend builds UI state in an isolated git worktree > teammate-backend scaffolds services and manages Redis distributed locks > teammate-adversary runs property-based fuzzing and catches memory leaks > direct inter-agent messaging eliminates central coordinator latency > tasks.json enforces atomic file mutexes across all active worktrees Inject this configuration prompt directly into Claude Code: "Configure my Claude Code workspace for autonomous agent teams: 1. Audit ~/.claude/teams and .claude/teams for teammate profiles fitting ux, backend, and adversary > Generate definitions only for missing roles > Pin model: sonnet, effort: medium for workers, and model: fable for adversary > Assign each worker an isolated git worktree branch to eliminate file conflicts 2. Enable agent teams in ~/.claude/settings.json: > Set CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 > Lock main session effortLevel to high 3. Initialize tasks.json in repo root with atomic file locking, allowing teammates to claim subtasks and resolve contracts directly over inter-agent messaging without coordinator intervention 4. Add a team rule to ~/.claude/CLAUDE.md: teammates must request an adversary security review before opening a pull request for the lead session Show every configuration diff first. Do not apply edits until confirmed" ↳ http://code.claude.com/docs/en/agent-teams

这条是引用着另一条写的：x.com/damkina7/status/2102860956724335039