recommend-ruleslisted

Recommend a tailored rule set for this project. Two-layer process: deterministic scoring of canonical rules (`bin/recommend-rules.sh`), then optional web search for stack-specific community rules. ## When to invoke - After `/ai:setup` on a brownfield repo where the default universal rule set feels incomplete. - When the user asks "which rules apply here", "find rules for this stack", "is there a Laravel rule pack?", etc. - When a previous emit was generic and the user wants to layer specificity. If the user wants only the deterministic ranking and no web step, run the CLI directly and stop after Phase 1. ## Process ### Phase 1 — Canonical scoring Run the deterministic scorer: ```bash $AI_KIT_ROOT/bin/recommend-rules.sh "$PWD" --json ``` It scores every rule in `$AI_KIT_ROOT/standards/rules/` against: - **Frameworks** (from `detect-tooling.sh`) — match against rule's `applies_to.frameworks`. - **Architectures** (frontend/backend detected) — match against `applies_to.architectures`. - **Repo age** (years since first commit) — must meet `repo_age_min_years`. - **Weight** (`high|medium|low`) — base score for universal rules. Present the result as a table to the user. Group by relevance tier: - **Strong fit** (score ≥ 3) — recommend enabling. - **Moderate fit** (score 1–2) — propose, but let the user opt out. - **Filtered out** (score 0) — show only if user asks "why isn't X here". Within each tier, surface the **feedback bucket** (rules whose source path is `standards