discuss

To capture user decisions for gray-area tasks, this Skill identifies scoping ambiguacies across UI, API, and data tasks, then uses AskUserQuestion prompts to collect your choices and generate a structured CONTEXT.md file.

The best way to stop downstream agents from re-asking questions is to seed a structured CONTEXT.md file with your captured decisions, enabling consistent execution without re-asking.

You structure context files by generating a global `.claude/CONTEXT.md` or a task-specific `.claude/context/{task-id}.md` file, ensuring downstream agents can retrieve the exact gray-area decisions needed for that workflow.

Yes, you need a TASKS.md file to identify the current task, which allows the Skill to pinpoint gray areas, collect decisions via prompts, and generate the corresponding context files.

Yes, you can automate context management by generating individual `.claude/context/{task-id}.md` files for each task, allowing multiple downstream agents to work concurrently with isolated, captured decisions.

The limitation of using CONTEXT.md is that it primarily targets gray-area scoping across UI, API, and data tasks, meaning highly deterministic workflows may not benefit from the decision-capture process.