← ClaudeAtlas

diagram-codebaselisted

Create self-contained HTML diagrams of systems, architectures, workflows, and code relationships. Use when the user asks for a diagram, an architecture picture, or a workflow visualization, or wants to communicate a design visually.
akshay-diwadkar/skills · ★ 2 · AI & Automation · score 72
Install: claude install-skill akshay-diwadkar/skills
# Diagram Codebase ## Purpose and authority Ground a shared model, obtain approval, then build one self-contained HTML diagram with the bundled template. Default to `narrative-architecture`. Minimum payload: title, fidelity, nodes, labeled edges, and agent metadata. Use `exact-code-graph` only for code-symbol review and `executive-concept-map` only for business ownership and outcomes. Do not create, overwrite, save, or verify diagram files in conversational Plan Mode. Model approval, directory creation, overwrite permission, and execution authority are separate decisions. Never hand-code an alternate renderer for normal output. ## Start After approval and outside Plan Mode, use absolute paths: ```bash python /absolute/skill-root/scripts/cli.py --repo-root /absolute/repo \ --input fidelity=narrative-architecture --input data=/absolute/payload.json \ --input output=/absolute/diagram.html --input create_dirs=no \ --input overwrite=no --format json run ``` Run each returned `next_command.argv` with its returned `cwd`. Read only `required_reads`, write only `allowed_writes`, and stop on every `blocking_reason`. Never change `create_dirs` or `overwrite` to `yes` without the matching user permission. ## Next-step loop Inspect repository evidence before asking questions. Follow the phased [Diagram Output Guide](references/diagram-output-guide.md) until purpose, audience, fidelity, entities, relationships, omissions, evidence policy, output path, and verification are co