[rig-summarizer] Daily file summary — 2026-10-12
メンテナーはふだん 1 日以内に返信
まだ誰も着手していません。
評価
- 難易度
- 5/5
- 見積もり時間
- 1週間以上
- 初心者へのやさしさ
- 1/100
- issue の種類
- ドキュメント
- 明瞭さ
- 説明が足りない
- 活発さ
- 活発
- 技術スタック
- typescript
- 領域
- tooling
調査の方向性
The issue contains a generated summary of Rig files, but it does not specify a change to make or a result to deliver. Start by reading .agents/skills/rig/rig.ts and the listed engine files to understand the project; the issue provides no completion criteria.
索引モデルが issue の本文から書いたものです。
説明
Global Summary
The project is a minimal TypeScript multi-agent harness ("Rig") centered on skills/rig/rig.ts, which provides schema-typed agents, prompt intents, sub-agent delegation, and workflow orchestration atop the Copilot SDK. Pluggable engine adapters (Anthropic, Codex, DeepSeek, Gemini, Pi) in skills/rig/engines/ share lifecycle and tool-schema helpers from utils.ts, letting the same agent API run against different model providers. Supporting infrastructure includes a harness-owned launch tool (launch-tool.ts/run.ts) that runs trusted Rig programs without exposing SDK credentials to arbitrary shell code, GitHub Actions drivers (.github/drivers/) that wire Codex/Pi sessions into the harness's permission and tool systems, and fixture/integration tests (.github/fixtures/, scripts/*.integration.test.ts) that validate a three-judge approval workflow and real single-agent sample runs across providers.
Individual File Summaries
| File | Summary |
|---|---|
.agents/skills/rig/engines/anthropic.ts |
Implements anthropicEngine, an AgentFactory wrapping the Anthropic SDK's beta tool runner with structured JSON output support and tool/text message handling. |
.agents/skills/rig/engines/codex.ts |
Implements codexEngine, wrapping @openai/codex-sdk threads to run prompts with optional structured output schemas; does not support Rig tools. |
.agents/skills/rig/engines/deepseek.ts |
Implements deepseekEngine using the DeepSeek Harness SDK session API, handling cancellation, cleanup, and completed-turn validation. |
.agents/skills/rig/engines/gemini.ts |
Implements geminiEngine, spawning the Gemini CLI as a subprocess per turn with session resumption and JSON output parsing. |
.agents/skills/rig/engines/pi.ts |
Implements piEngine, wrapping @earendil-works/pi-agent-core to run prompts and tools against a configurable Pi model registry. |
.agents/skills/rig/engines/utils.ts |
Shared engine utilities: a agentLifecycle helper enforcing single-turn concurrency with abort support, plus tool schema/result-text conversion helpers. |
.agents/skills/rig/globals.ts |
Exposes ambient call, pipeline, and parallel helpers that delegate to the active WorkflowContext, simplifying workflow-style agent invocation. |
.agents/skills/rig/launch-tool.ts |
Defines createRigLaunchTool, a Copilot SDK tool that spawns an isolated Rig launcher subprocess to run trusted TypeScript source without exposing connection credentials to it. |
.agents/skills/rig/rig.ts |
The core Rig runtime: declarative schema helpers, prompt intents, agent/workflow construction, Copilot SDK engine, and the launcher CLI entry point. |
.agents/skills/rig/run.ts |
Thin entry point that imports and runs runLauncherCli from rig.ts when executed as the main module. |
.github/drivers/codex-fixture-mcp.ts |
Implements a Docker/chroot-isolated Codex fixture launcher and a minimal JSON-RPC MCP server exposing a single run_rig tool for the three-judge integration test. |
.github/drivers/copilot-sdk-driver.ts |
A GitHub Actions driver that creates a Copilot SDK session, registers the harness-owned Rig launch tool, and enforces single-use execution of a prevalidated fixture. |
.github/drivers/pi-rig-extension.ts |
Registers a run_rig Pi agent tool that launches the checked-in three-judge fixture exactly once and validates/writes its JSON result. |
.github/fixtures/assert-three-judges.ts |
Provides assertThreeJudges, validating the shape and verdict of a three-criterion (clarity/safety/feasibility) judgment result, plus a CLI entry point for ad hoc checks. |
.github/fixtures/pi-gateway.ts |
Builds a Pi model registry from a harness-provisioned "aw-gateway" provider configuration, validating its shape and exposing the selected model. |
.github/fixtures/provider-fixture.ts |
Selects and configures a provider engine (Codex, Gemini, DeepSeek, or Pi) based on RIG_JUDGE_ENGINE, returning a configured three-judge workflow. |
.github/fixtures/provider-three-judges.ts |
Minimal entry point that default-exports the workflow created by createProviderFixture(). |
.github/fixtures/three-judges.ts |
Defines the reusable three-judge workflow (clarity/safety/feasibility) used to sanity-check provider engine integrations with a dummy request. |
scripts/haiku.integration.test.ts |
Integration tests that spawn the Rig launcher against real haiku/sonnet/complex samples using live Copilot SDK credentials when available. |
scripts/rig-launch-tool.integration.test.ts |
An opt-in live integration test verifying the harness-owned run_rig launch tool runs the three-judge fixture once without exposing connection credentials. |
Generated by Daily Rig File Summarizer · copilot · auto · 94.7 AIC · ⌖ 6.22 AIC · ⊞ 7.5K · ◷
- expires on Oct 19, 2026, 2:51 PM UTC
- 主要言語
- TypeScript
- スター
- 14
- フォーク
- 0
- 平均マージ
- 3時間 38分
- マージ済み PR(30日)
- 29
環境構築
このプロジェクトには開発コンテナ、Dockerfile、コントリビューションガイドがありません。まず README を読み、一般的な手順ははじめてのコントリビューションガイドを参照してください。
はじめの一歩
- issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
- 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
- リポジトリをフォークし、ブランチを切って変更します。
- issue 番号を参照したプルリクエストを送ります。
githubnext/rig のほかの issue
-
agentic-workflows
難易度 3/5 半日 初心者へのやさしさ 15/100
githubnext/rig#643 ·
メンテナーはふだん 1 日以内に返信
-
agentic-workflows
難易度 4/5 3〜5日 初心者へのやさしさ 15/100
githubnext/rig#642 ·
メンテナーはふだん 1 日以内に返信
-
agentic-workflows
難易度 3/5 半日 初心者へのやさしさ 22/100
githubnext/rig#641 ·
メンテナーはふだん 1 日以内に返信
-
ai-agent automation
難易度 4/5 3〜5日 初心者へのやさしさ 12/100
githubnext/rig#640 ·
メンテナーはふだん 1 日以内に返信
-
ai-agent automation
難易度 3/5 1〜3時間 初心者へのやさしさ 8/100
githubnext/rig#638 ·
メンテナーはふだん 1 日以内に返信
似ている issue
-
難易度 2/5 1〜3時間 初心者へのやさしさ 62/100
one-zero-eight/apply-bot#39 ·
-
難易度 2/5 1〜3時間 初心者へのやさしさ 72/100
bigbluebutton/bbb-ui-components-react#108 ·
メンテナーはふだん 2 日以内に返信
-
難易度 2/5 1〜3時間 初心者へのやさしさ 82/100
メンテナーはふだん 1 日以内に返信
-
難易度 2/5 1〜3時間 初心者へのやさしさ 70/100
メンテナーはふだん 1 日以内に返信
-
high priority
難易度 2/5 1〜3時間 初心者へのやさしさ 62/100
CI-CMG/deep-sea-corals-exb#84 · コメント 1 件 ·