Interviews users in structured rounds to stress-test plans against domain glossaries and ADRs.
npx skills add https://github.com/reddb-io/red-skills --skill start-reddb-io
Please help me install this Agent Skill. Skill: start Source: https://github.com/reddb-io/red-skills/tree/main/plugins/dev/skills/engineering/start Command: npx skills add https://github.com/reddb-io/red-skills --skill start-reddb-io
.red/CONTEXT.md glossaries the moment a term is settled, and offers ADRs only when a decision is hard to reverse, surprising, and a real trade-off. - External reference ingestion: Fetches URLs, PDFs, and files through a wiki cache with single-line receipts, and lands changed docs through a worktree-based PR finalizer at session end. - Use Case: Before starting a billing feature, paste your plan and let the session expose that your code cancels entire Orders while your plan assumes partial cancellation, then record the resolved terms and an ADR. ## Quick Start Ask the AI to grill your plan by invoking the start skill with a pasted plan, URL, or file path as the argument.High-intent search queries and answers about installing and using this skill.
Invoke the start skill with your plan as prose, a URL, or a file path. It interviews you in numbered rounds, asking only questions whose prerequisites are settled, and challenges contradictions between your statements, the code, and existing documentation.
An ADR records that a decision was made and why, stored as a sequentially numbered file in .red/adr/. Write one only when the decision is hard to reverse, surprising without context, and the result of a real trade-off between genuine alternatives.
Keep a .red/CONTEXT.md file with one-sentence definitions, aliases to avoid, relationships, and flagged ambiguities. The skill updates it inline the moment a term is resolved during the interview, one term per edit.
Yes, URLs and file paths introduced with intent to ingest are fetched through the wiki skill and cached under .red/wiki/raw/ for reuse across sessions. If caching is declined, content is read directly into context and the receipt is marked not cached.
No, the skill is read-only except for documentation updates. It ends in shared understanding, explicitly refusing to implement code, propose a final plan, or generate a spec document.
SebasArenClassify raw materials and generate wiki pages from PDFs, notes, and URLs.
ThorStarlordConverts current conversation and repository understanding into a complete PRD and publishes it to the issue tracker with ready-for-agent label.
joshka0Save key learnings, decisions, and gotchas into foxctl memory with typed classification.
prabaljainnInterviews the user question-by-question to stress-test plans and design decisions.