issue-resolve_skill

Overview

This skill provides a unified issue resolution pipeline that creates, converts, and queues actionable solutions from multiple sources. It routes input to one phase (plan, convert, brainstorm, or queue), runs AI-driven planning or subagents, and can export tasks as .task/TASK-*.json for execution. The design favors CLI-first operations and supports an auto mode to skip prompts.

How this skill works

Input and flags are parsed and an auto-detection routine selects the source (issue IDs → plan, SESSION= → brainstorm, file paths → convert, or explicit --source). The orchestrator reads only the selected phase document, spawns subagents for planning or conversion, waits for results, binds solutions to issues, optionally exports tasks, and suggests next steps. All issue and solution CRUD uses the ccw issue CLI; files are not read directly for operational data.

When to use it

• Plan and batch AI-generated solutions for one or more issue IDs (Explore & Plan).
• Convert an existing lite-plan, workflow session, or markdown artifact into bound issue solutions.
• Translate brainstorm session outputs into concrete issues with solutions.
• Form ordered execution queues from bound solutions before /issue:execute.
• Export solution tasks to .task/TASK-*.json for downstream task orchestration or CI.

Best practices

• Always prefer the ccw issue CLI for list/read/update/bind operations; avoid reading large issue files directly.
• Use --yes (-y) in automation contexts to default to Explore & Plan without prompts.
• Keep batch-size small (default 3) to limit agent context and improve result quality.
• Export tasks only when solutions are complete; verify exported TASK-*.json against task-schema.json.
• Close subagents after wait completes to free resources and avoid zombie tasks.

Example use cases

• Auto-plan GH-123,GH-124 with codex prompt to generate bound solutions and task exports.
• Convert .workflow/.lite-plan/my-plan.md into issue solutions and append tasks to an existing solution with --supplement.
• Turn SESSION="BS-rate-limiting" brainstorm results into issues and solutions, then form a queue.
• Form a single or multi-queue execution plan (--queues) from planned solutions and hand off to /issue:execute.
• Run in CI: codex -p "<ids>" -y --export-tasks to produce .task/TASK-*.json and a plan.json artifact.