Rights Manager string skill
The problem
The Rights Manager engineering team was writing product strings without CD guidance—generic language, wrong terminology, no awareness of legal nuance. Every string that landed in my review queue was starting from scratch, because the engineers writing them had no guidance to start from. I was reviewing and explaining the same issues repeatedly. I was the bottleneck.
What I built
A Claude skill integrated into the team's existing plugin. Three modes:
Draft
Generates first-pass strings using RM terminology rules, Meta's style guide, and Blueprint component patterns. Engineers start from something already in the right language.
Audit
Validates existing strings against the same standards. Catches wrong terminology, style violations, and pattern mismatches before they reach CD review.
Lookup
Answers terminology questions directly. "Is it 'asset' or 'content'?" "What's the correct term for a rights holder?" No waiting for CD availability.
The skill automatically loads my terminology source-of-truth document, so the guidance it operates from is always current. The document is the single source of truth for RM content standards—the same one I review against manually.
How I designed the limits
I designed what the skill can't do as carefully as what it can. It explicitly states it doesn't handle Tier 2 or Tier 3 decisions, can't assess legal risk, and doesn't manage localization.
That's not an oversight—that's how you make a tool trustworthy. If engineers don't know what to hand off to CD review, they won't hand anything off. Stating the limits is what keeps the skill from becoming a reason to skip the review process it was built to support. That's systems maturity.
How it shipped
I participated in the code review as a technical contributor, not just a reviewer. During review, I caught that key behavioral guardrails had been trimmed during integration—constraints that prevented the skill from operating outside its defined scope. I pushed to get them restored before the diff landed.
The outcome
The full engineering team got access. The work got recognized at the org level and was integrated into the team's engineering plugin.
“Rebecca and I worked closely together during her time at Meta; she was the team's Content Designer and I was her engineering counterpart. She regularly went above and beyond the ‘content’ scope—she contributed significantly to the product direction and helped shape the strategic vision long-term. The product we collaborated on is deeply complicated and has been built up over years without a cohesive content and design strategy; she stepped up to the challenge of disentangling that complexity and delivered concrete proposals that addressed those challenges, which our customers have greatly appreciated. As part of this, Rebecca partnered closely with engineering, design, and product management during her time on the team, communicating regularly and working with the team to drive her proposals forward. Rebecca also embraced the company vision and developed AI tooling internally that helped her and the engineering team move more efficiently. I really valued working with Rebecca and would do so again in a heartbeat. She will be a strong contributor no matter where she goes.”
Artifacts
Interactive demo—coming once terminology doc is available