agentdev-case-run-execution-adapter

Defines the delegation protocol connecting case-run issue execution to execution subagents.

Updated Mar 7, 2026
One-click install
npx skills add https://github.com/yogata/agent-dev-flow --skill agentdev-case-run-execution-adapter-yogata
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: agentdev-case-run-execution-adapter
Source: https://github.com/yogata/agent-dev-flow/tree/main/src/opencode/skills/agentdev-case-run-execution-adapter
Command: npx skills add https://github.com/yogata/agent-dev-flow --skill agentdev-case-run-execution-adapter-yogata

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? When an AI-driven development workflow (case-run) needs to hand off a single GitHub Issue's implementation to an execution subagent, there must be a strict contract governing how work is delegated, how results are returned, and how failures are recovered. This Skill defines that adapter protocol so delegated implementation stays isolated, auditable, and recoverable. ## Core Features & Use Cases - Four-state result contract: Execution subagents return exactly one of completed-pr (with PR number/URL), blocked, failed, or delegation-unavailable, with PR bodies and Issue comments as the single source of truth. - Worktree isolation enforcement: All file edits are confined to the delegated worktree root (.worktrees/{N}-{type}/), with self-verification helpers and immediate failed status on violation. - Recovery and quality gates: Defines a 3-point receipt gate (result state, commit hash, PR URL), background delegation loss recovery via durable state, test-fix loops per test strategy item, and optional adversarial-review integration before implementation. - Use Case: A case-run orchestrator delegates Issue #980 to an execution subagent with a structured delegation prompt; the subagent implements, runs quality gates, creates a PR via agentdev_gh, and returns completed-pr with the PR URL. ## Quick Start Load this adapter skill when delegating a case-run Issue implementation to an execution subagent and follow its result contract and worktree isolation rules.

Frequently Asked Questions about agentdev-case-run-execution-adapter

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I delegate GitHub Issue implementation to a subagent?▼

The case-run orchestrator launches an execution subagent that loads this adapter skill, passing a delegation prompt containing the execution command, worktree root, branch name, and Issue body. The subagent implements, runs quality gates, creates a PR, and returns a structured result.

What result states does the execution adapter return?▼

The adapter returns exactly one of four states: completed-pr (with PR number and URL), blocked (answerable blocker, details in Issue comment), failed (unanswerable blocker, structured Issue comment), or delegation-unavailable (infrastructure could not launch, Issue returns to pending).

How does the adapter enforce worktree isolation?▼

The execution subagent may only edit files under the delegated worktree root (.worktrees/{N}-{type}/). It self-verifies its location before starting, and if it detects work in the main repository it stops immediately and returns failed with a structured Issue comment.

What happens when a delegated subagent crashes or disappears?▼

The orchestrator recovers using durable state only: worktree git status, commit presence, PR existence, and Issue comments. Unstarted work is re-delegated synchronously; interrupted work with residual changes is classified, verified, and either committed into a PR or marked blocked/failed.

When should I not use this execution adapter skill?▼

Do not use it for workflow state management, Epic or multi-Issue orchestration, or Issue completion checkbox evaluation. Those responsibilities belong to case-run, case-auto, and case-close respectively; this skill only defines the delegation contract.