← ClaudeAtlas

interactive-codebase-atlaslisted

Explicitly-invoked only. Create, update, audit, or explain an interactive, animated, ADHD-friendly browser-based "Codebase Atlas" that teaches how any repository works. Invoke only when the user explicitly names this skill or its slash command.
Martin-Kocev/skills · ★ 0 · AI & Automation · score 67
Install: claude install-skill Martin-Kocev/skills
# Interactive Codebase Atlas Turn an unfamiliar repository into a trustworthy island archipelago: an explorable technical museum that builds a working mental model instead of listing files. Let the adventure metaphor support comprehension; never let it replace evidence. ## Activation **Require explicit invocation.** Run this skill only when the user calls `/interactive-codebase-atlas`, names the skill, or points at an existing atlas directory and asks to update, audit, or explain it. Do not activate for a generic request to document, map, or explain code. ## Non-negotiable safety **Never let the atlas enter the repository's Git history.** Default to the sibling directory `../<repo-name>-codebase-atlas/`. Never add, commit, push, or include atlas files in a PR. Never modify production source, schema, config, tests, or infrastructure to support the atlas. Read `references/safety.md` before writing the first file on every run; it outranks all visual and workflow guidance. ## Execution contract 1. Resolve mode and inputs by discovery. 2. Read `references/safety.md`; establish the output location and Git isolation. 3. Run `scripts/git-metadata.mjs` and record the exact repository state. 4. Read the selected mode playbook plus its linked references; follow them in order. 5. Validate with `references/validation.md`. 6. Report with `references/report.md`. Mark a checklist item N/A only when repository evidence proves the capability absent. Never mark it N/A because it was inc