What problem does it solve? Setting up per-layer specialist agents and engineering-standards skills by hand duplicates information already declared in a repo's CLAUDE.md, and hand-written copies drift out of sync with the declared layer chain. This Skill reads the chain declaration once and generates the matching agent and skill files, creating only what is missing and never overwriting existing files. ## Core Features & Use Cases - Chain-driven generation: Reads the layer chain, build/test commands, and contracts from the repo's own CLAUDE.md and produces one layer specialist, one slice-mode variant, and one standards skill per layer. - Idempotent, non-destructive runs: Skips files that already exist, detects stale model comments and unfilled placeholders, and reports disagreements between generated files and the current chain without editing anything. - Structured reporting: Ends every run with a per-file report classifying each row as created or skipped, plus findings like still-a-scaffold or disagrees-with-CLAUDE.md, and flags when a session restart is needed. - Use Case: After scaffolding a new repo whose CLAUDE.md declares a schema-api-web chain, run this Skill to populate .claude/agents/ and .claude/skills/ with the nine corresponding files, then re-run it months later when a fourth layer is added to generate only that layer's files. ## Quick Start Ask the AI to adapt this repository to its declared stack by generating the layer specialists and standards skills from the repo's CLAUDE.md.
