Brand

A person's brand is spread across a dozen public surfaces that drift out of sync: old project names, stale focus statements, dead links, outdated writing lists. This skill is the one place that knows every surface, how each one updates, and what "current" means, so a single BrandUpdate run brings them all back into line.

Source of truth for what the brand currently is: the files listed under ## Source-of-truth reads in Surfaces.md (focus and missions, what is live, bio facts) plus the freshest posts from the blog feed ($BLOG_FEED). Never duplicate those facts into this skill; read them at run time.

Ideal State

• Every surface a run touches says what the principal is focused on now, as read from the USER inventory and source-of-truth files at run time, never from memory or this skill.
• Nothing goes public under the principal's name before they have seen the rendered result and approved that one surface.
• Every run ends with a table in its run ISA naming each surface or idea, what changed and the evidence, including what was left unswept or rejected.
• A missing or invalid Surfaces.md stops the run at Check.ts with a pointer to Setup.md; no surface is ever swept from guesses.

Intent Routing

Intent	Trigger	File
BrandUpdate	"brand update", "brand sweep", "update my GitHub profile", "refresh my bios/profiles"	Intents/BrandUpdate.md
UpdateIdeasPage	"update my ideas page", "refresh the ideas page", "parse my latest ideas"	Intents/UpdateIdeasPage.md

File	Read when
Setup.md	first-time setup, or Check.ts exits non-zero

Data & Config

Every personal fact (the public-surface inventory with URLs, repo paths, handles and update mechanisms; the ideas-page location and structure; the blog and feed facts $BLOG_URL, $BLOG_FEED, $BLOG_REPO; per-surface rules) lives in <configRoot>/LIFEOS/USER/CUSTOMIZATIONS/SKILLS/Brand/Surfaces.md (configRoot is $CLAUDE_CONFIG_DIR when set, else ~/.claude). This skill body is generic.

Both intents start with bun ~/.claude/skills/Brand/Tools/Check.ts. It validates Surfaces.md and, when the file or a required section is missing, exits non-zero and points at the guide. First-time setup: Setup.md.

Gotchas

• Use the feed URL recorded in Surfaces.md ($BLOG_FEED). A blog's feed often lives somewhere other than the obvious paths (/feed.xml, /rss.xml, /atom.xml), so confirm the working one with curl before recording it.

Examples

Example 1: Full brand sweep

User: "Run a brand update — everything feels stale"
→ BrandUpdate intent, full sweep
→ A drift table in the run ISA covers every inventory row; each fix went live only after the owner approved that surface

Example 2: Single surface

User: "Update my GitHub front page based on recent focus"
→ BrandUpdate intent, scoped to the GitHub profile surface
→ The README is pushed after approval of its GitHub render; every other surface is listed as unswept

Example 3: Ideas page

User: "Refresh the ideas page with my latest ideas"
→ UpdateIdeasPage intent
→ New principal-original ideas are on the live page, verified by screenshot after an approved deploy; rejected captures are listed with reasons