discover-brand

An orchestration skill for autonomous discovery of brand materials across connected enterprise platforms, including Notion, Confluence, Google Drive, Box, Microsoft 365 (SharePoint/OneDrive), Figma, Gong, Granola, and Slack. It is triggered by requests like discovering brand materials, finding a style guide, auditing brand content, or asking where brand documents live. The skill coordinates a separate discover-brand agent that searches the connected platforms, triages sources, and produces a structured discovery report with open questions, then optionally feeds that report into brand-voice guideline generation.

The workflow begins by orienting the user with a plain explanation of the four stages — search, analyze, generate guidelines, and save — and waiting for confirmation. It reads optional settings from .claude/brand-voice.local.md (company name, enabled platforms, search depth, max sources, and known material locations), then validates platform coverage by classifying connected platforms into document platforms, where formal brand guidelines usually live (Notion, Confluence, Google Drive, Box, Microsoft 365), and supplementary platforms valuable for patterns but unlikely to hold brand docs (Slack, Gong, Granola, Figma). Coverage rules are explicit: if no document platform is connected it stops and tells the user to connect one; if none of Google Drive, Microsoft 365, or Box is connected it warns about likely gaps but proceeds; and if only one platform is connected it warns that cross-source validation and confidence will be lower.

After confirming scope with a single brief question, it delegates to the discover-brand agent via the Task tool, passing company name, enabled platforms, search depth, and any known locations. The agent runs a four-phase algorithm — broad parallel discovery, source triage and ranking, deep fetch from top sources, and a discovery report with open questions — and the skill then presents the report, summarizing sources found, key brand elements, conflicts, and questions needing team input. Open questions (such as a 2023 style guide conflicting with a 2024 update, or missing social-media guidelines) always include an agent recommendation and are framed as confirm-or-override. Next steps include generating guidelines immediately, resolving open questions first, saving the report, or expanding the search; guidelines are written to .claude/brand-voice-guidelines.md only after approval. Detailed search and ranking logic live in references/search-strategies.md and references/source-ranking.md.