poorvith-mp/skills-agents

agent-architecture

Design role decomposition, message passing, shared state and failure isolation across agents.

ソースを見る

使い方

  1. インストールコマンドをコピーし、自分の環境へ導入する前にソースを確認します。
  2. 必要なツール、認証情報、対応環境を原文で確認します。
  3. 入力と成果物を指定し、出力を資料と照合します。
npx skills add https://github.com/poorvith-mp/skills-agents --skill agent-architecture

開始前の確認

  • 導入にはモデル契約、API枠、ファイルへのアクセス許可は含まれません。
  • スター数や導入数は活動の指標であり、独立した品質・安全性試験ではありません。
リポジトリの原文

見出し、例、コード、表、リンク、参照画像を含む原文を表示しています。

Agent Architecture

Design multi-agent systems where coordination does not become the failure mode.

Process

  1. Justify the second agent. Most tasks that look multi-agent are one agent with better tools. Split only when roles need genuinely different context or run concurrently.
  2. Decompose by context boundary, not by job title. An agent exists to hold a distinct working set, not to mirror an org chart.
  3. Define the contract between agents — the exact shape of what is passed, and what is guaranteed about it. Free-text handoff is where multi-agent systems lose information.
  4. Decide where state lives. Shared mutable state between agents needs the same care as between threads.
  5. Isolate failure. One agent failing or looping must not stall the others; set timeouts and a supervisor that can terminate.
  6. Make the trace reconstructable. When output is wrong, you need to see which agent decided what, with what input.
  7. Bound the total cost. Agents that can spawn agents need a hard ceiling.

Deliverables

  • Role decomposition with the context boundary justifying each agent
  • Message contracts between agents
  • Failure isolation, timeouts and supervision model
  • Trace format and cost ceiling

Verification & Quality Checklist

  • [ ] Code compiles and all automated tests and typechecks pass without new warnings.
  • [ ] Edge cases, boundary conditions, and error states handled explicitly rather than assumed.
  • [ ] No hardcoded secrets, credentials, or insecure defaults introduced.
  • [ ] Changes are covered by a test that fails without them.

Anti-Patterns & Constraints

  • NEVER weaken or skip a failing test to make a change land.
  • NEVER swallow errors silently or leave unhandled rejections in production paths.
  • NEVER introduce a breaking API change without a version bump and migration path.