Brand Unification Skill — AI System Brief | Maneesh Maddala Skip to main content

Brand Unification Skill

From a public URL to a scored brand-migration audit with a routed escalation.

Claude Skills · Progressive Disclosure · Claude Enterprise · Rubric Scoring · Prompt Engineering · Guardrails · Governance

My company was consolidating several legacy brands into one. Thousands of public pages still named the retired brands: blog posts, articles, research papers, campaign assets. Content writers and creative services had no consistent way to find every reference, judge how far off-brand a page was, or decide what needed legal review.

To address it, I proposed a Claude skill and built it: content and creative teams import it, hand it a URL, and it audits the page against the unified brand’s rules. Every finding traces to a written rule, and the skill rewrites and recommends but never publishes.

A user in Claude Desktop pastes a URL and asks Claude to audit it. That request loads the skill, which pulls the page and, following progressive disclosure, reads only the rule files it needs for the current step, so the context stays small and uncorrupted across a long audit.

The skill runs the content through four ordered rule sets: brand remediation, tone and voice, content rewrite, and escalation. Each set scores the content against its rubric and records the specific rule behind every finding.

It returns a score per rubric, a rewritten draft of the off-brand passages, and an escalation call: keep it with the content team, or send it to legal. The writer applies the changes; the skill never publishes.

Architecture diagram
Brand unification skill audit flow A user in Claude Desktop pastes a URL, document, PDF, or deck and asks Claude to audit it. That loads the brand skill, which uses progressive disclosure to read only the rule files each step needs. The content passes through four ordered rule sets — brand remediation, tone and voice, content rewrite, and escalation — each scoring it against a rubric. The skill returns an audit report with a score per rubric and a rewritten draft, then routes the finding to the content team for in-scope fixes or to legal when it needs review. The draft returns to the writer, who makes the change; the skill never publishes. SKILL · RULES + RUBRICS audit run rules in scope needs review draft to the writer · skill never publishes Claude Desktop URL · doc · PDF · deck Brand skill progressive disclosure Brand remediation Tone & voice Content rewrite Escalation Audit report score per rubric + rewritten draft Content team creative services Legal brand & IP audit flow draft back to the writer skill · rules + rubrics

A skill with broad rewrite instructions can drift into editorializing or changing meaning. Every pass is bound to a written rule set, and the skill records the specific rule behind each finding, so a reviewer can check any change against its source.

A long audit can fill the context with rule text and stale page content until the model loses the thread. The skill uses progressive disclosure: it loads only the rule files each step needs and drops page content once a step is scored, so the working context stays small.

Some off-brand content is a legal question, not a copy edit. The escalation rule set routes each finding to the content team or legal, and the skill returns a score and a draft but never publishes, so a person still makes the edit and the release call.

The brand migration finished in about a month against a three-month estimate: one week to build and test the skill, three weeks to work through the content.

Content writers and creative services run the audit themselves in Claude, with no dependency on a central brand reviewer.

What started as a web-page tool now audits Word documents, PDFs, and presentations, and is the organization’s default path for content migration.