skills.rest
yogata
yogataCOMMUNITY
Explore all skills by this author →
GitHub

agentdev-workflow-case-auto

Orchestrates autonomous case-open through case-close workflows with stop detection and conflict resolution.

Updated Mar 7, 2026
Search Tags:#pull request#workflow orchestration#issue management#conflict resolution#github automation#stop detection#autonomous pipeline
One-click install
npx skills add https://github.com/yogata/agent-dev-flow --skill agentdev-workflow-case-auto-yogata
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: agentdev-workflow-case-auto
Source: https://github.com/yogata/agent-dev-flow/tree/main/src/opencode/skills/agentdev-workflow-case-auto
Command: npx skills add https://github.com/yogata/agent-dev-flow --skill agentdev-workflow-case-auto-yogata

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Running a full development case from requirement draft to merged PR requires manually chaining multiple commands (case-open, case-ready, case-run, case-close) and handling stops, conflicts, and resume points by hand. This Skill automates that end-to-end orchestration while enforcing guardrails and stop conditions. ## Core Features & Use Cases - Autonomous pipeline orchestration: Runs case-open → case-ready → case-run → case-close (with a case-revise exception path) from a requirement draft or GitHub Issue number, including parallel case-run execution via background tasks (up to 5 concurrent). - Stop detection and decision resolution: Detects 11 stop conditions, classifies stop reasons across 9 axes, propagates adversarial-review user-decision-required signals, and performs bounded parent decision resolution to autonomously answer or escalate. - Conflict resolution and reporting: Handles PR merge conflict escalation Levels 2/3 (inline case-run re-execution, merge order changes) and produces completion reports with L1 timestamps and four-dimensional result aggregation. - Use Case: Point it at a requirement draft in .agentdev/drafts/ and it creates the GitHub Issue, finalizes the Definition Package, implements the change in parallel waves, merges the PR, and closes the Issue—stopping with a classified reason and resume command whenever human judgment is required. ## Quick Start Invoke the /agentdev/case-auto command with a requirement draft path or GitHub Issue number to run the full autonomous case workflow.

Frequently Asked Questions about agentdev-workflow-case-auto

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

FAQPage Schema
How do I run a full development case automatically from a requirement draft?▼

Invoke the /agentdev/case-auto command with a draft path or no arguments to process all drafts in .agentdev/drafts/. The workflow runs case-open, case-ready, case-run, and case-close sequentially, creating the Issue, implementing the change, merging the PR, and closing the Issue.

How does case-auto handle GitHub Issue input instead of a draft?▼

When given an Issue number or URL, it reconstructs the Root Case durable state and resumes from the appropriate stage: case-ready if open, case-run if ready/running/review, or case-revise if re-agreed Definition changes exist. Closed Issues are reported as complete without re-execution.

What happens when case-auto hits a merge conflict?▼

Merge conflicts follow a three-level escalation: Level 1 (case-close) attempts mechanical rebase resolution, Level 2 re-runs inline case-run with conflict context up to 2 times, and Level 3 changes merge order or isolates blocked units. If all levels fail, execution stops with a classified reason.

When does case-auto stop and wait for user input?▼

It stops on 11 conditions including requirement deviation, out-of-repo changes, DB migrations, deploys, credential changes, unrecoverable CI failures, and unresolvable conflicts. It also stops when adversarial-review raises user-decision-required or when bounded decision resolution detects contradictions or new user judgment needs.

Can case-auto resume after a stop without conversation context?▼

Yes. Resume points are reconstructed from durable state: case_auto_started_at, L1 timestamps, Issue/PR existence, Epic Issue status tables, and Root Case state with resume_command. The stop report explicitly states the resume point and next command, so no conversation memory is required.

What operations are outside case-auto's autonomous scope?▼

Only changes persisted as files in the repository are autonomous targets. DB migrations, deploys, cloud resource operations, external SaaS configuration, billing, permissions, credentials, out-of-repo data changes, and notifications are excluded and trigger stop conditions.

Related Skills

Explore More Software Engineering Skills→
mattpocockmattpocock
242.8k

writing-shape

Shapes raw markdown material into a structured article paragraph by paragraph.

Community
Intermediate
sobhanashinesobhanashine
1

pr-manager

Automate pull request reviews, checks, merges, and issue closure via GitHub API.

Community
Advanced
lushly-devlushly-dev
4

manage-projects

Orchestrate GitHub Projects, issues, and parallel Claude agent waves for feature lifecycle management.

Official
Advanced
oscardlfroscardlfr
1

test-full-parallel

Run all test cases in parallel with coverage using Gradle.

Community
Advanced