brandlisted

# Brand Voice Analysis and Guidelines Generation ## Phase 0: Grounding Read `${CLAUDE_PLUGIN_ROOT}/references/grounding.md` and load any `_grounding/` folder it finds. Client documentation outranks every default in this skill, and any claim rules it contains bind the output. Name the loaded files at the top of what you produce. Resolve the business type via `${CLAUDE_PLUGIN_ROOT}/references/business-context.md` before assigning an archetype — the same voice reads as confident to one audience and unreliable to another. Read `${CLAUDE_PLUGIN_ROOT}/references/output-location.md`. Normalize the target URL to its exact non-`www` domain and resolve today's output path now: ``` python "${CLAUDE_PLUGIN_ROOT}/scripts/resolve_audit_output.py" --purpose BRAND-VOICE --scope <domain> --extension md ``` Use `python3` on macOS/Linux. Retain `audit_dir` (to check for a same-scope `MARKETKIT - COMPETITOR-REPORT - <domain>.md`) and the exact `output_path` for the final write. Then resolve optional report metadata from the same working directory: ``` python "${CLAUDE_PLUGIN_ROOT}/scripts/resolve_report_metadata.py" --toolkit "Market Context Kit" --host <exact active host> --provider <exact active LLM provider> --model <exact active model id> ``` Never guess a runtime value. Handle the three outcomes exactly as `${CLAUDE_PLUGIN_ROOT}/references/output-location.md` specifies: `null` means write no metadata block at all, a JSON object means reproduce its fields verbatim as YAML front matt