Infopanel · Claude Skills
You wished Claude would just remember. It can now.
If you use Claude daily for real projects — not one-off asks, but sustained creative and strategic work — you already know how powerful a thinking partner it can be. You also know the overhead is real. Re-establishing context, re-explaining who you are and how you work, correcting the same misunderstandings thread after thread. You get momentum back once the context is in place. But imagine if you never had to rebuild it at all.
That’s what Claude Skills makes possible. And we built three of them — right here, in a working session — to show you exactly how.
What Claude Skills actually is
A Skill is a reusable knowledge package you upload to Claude once. Claude loads it automatically — without you re-explaining anything — whenever it’s relevant to what you’re working on.
Think of it as the onboarding document you always wished you could hand a new collaborator and have them actually absorb. Except it works every single time, in every new thread.
Where to start: your biggest recurring friction points
We started with the context we felt like we’d explained a thousand times. Not project status — that changes. Not “we completed 3 pages, 4 more to go” — that belongs in handoff documents. The evergreen things. The things that are always true regardless of what we’re building that day.
The evergreen test
Before adding something to a Skill, ask: will this still be true in 6 months? If yes — it belongs in a Skill. If it depends on where a project currently stands — it belongs in a handoff document or project file.
The three Skills we started with — and why
Skill 1
Triad Context
Who we are, how we work, the methodology, the ecosystem, key sites and their purpose, and how to work with Maura. The foundational brief that every new Claude thread needs — but no one should have to write twice.
Replaces: the wasted hours rebuilding context from scratch at the start of every thread.
Skill 2
WordPress Constraints
Hard-learned deployment rules for three WordPress sites — what’s supported, what gets stripped, proven workarounds, and the exact deployment workflow. Every constraint discovered across 11 threads, in one place.
Replaces: re-discovering the same broken SVG or stripped style tag over and over again.
Skill 3
Brand Guidelines
Voice, tone, visual style, typography, color palette, and how the brand manifests differently across each platform. Including the things that are easy to miss: “we” not “you,” not afraid to be provocative, never preachy.
Replaces: the brand correction pass at the end of every content draft.
How Skills fits in The Human–AI Loop
The Human–AI Loop is built on a simple premise: the human’s energy belongs in ideation, judgment, and decision-making — not in administrative overhead. Every minute spent re-establishing context is a minute not spent on the work that actually matters.
Skills directly reduce the human overhead that the Loop was designed to eliminate. When Claude already knows who you are, how you work, and what constraints apply — you arrive at the creative and strategic work faster. The Loop’s Test → Build → Codify → Share stages all run more cleanly when the foundation is already in place.
Without Skills
Hours rebuilding context. Human energy spent on setup, not work.
With Skills
Context loads automatically. Human energy goes straight to ideation and judgment.
The insight nobody tells you: build Skills inside your best threads
The documented path for creating a Skill starts from scratch: open Skill Creator, describe what you want, Claude generates a template. That works well for simple, self-contained workflows.
But for complex, context-rich Skills — the kind that capture how you actually work — there’s a better approach. Build them inside the threads where that context already lives.
The approach that worked for us:
Stay in the working thread — the one with hours of real context already established
Ask Claude to write the Skill content directly — it already knows the context, constraints, and patterns
Review, refine, and add what’s missing — you’ll know immediately what feels off
Save the skill directly from the preview — no zip, no upload needed when building from the interface
The frame that makes this click
When context is the Skill — and when it isn’t
The blank-thread UX makes perfect sense once you understand that there are two fundamentally different kinds of Skills. Knowing which kind you’re building tells you exactly where to build it.
Build in a rich thread
When context IS the Skill
Sometimes what you need Claude to know is the collaboration itself — who you are, how your team works, what constraints you’ve learned through practice. That knowledge doesn’t exist in a vacuum. It emerged from real work, and it lives in the threads where it was built.
You can’t write this kind of Skill accurately from a blank window. The thread is where the context lives — so the thread is where the Skill gets built.
Examples
Team context and operating patterns · Brand voice and tone · Technical constraints learned through practice · Methodology and collaboration architecture
Blank thread works fine
When context ISN’T the Skill
Sometimes what you need is a self-contained behavior rule — one that doesn’t depend on who you are or how you work. A developer who wants Claude to always write Python with type hints, always return JSON, always follow REST naming conventions can write that Skill from scratch in 60 seconds.
The blank-thread UX is optimized for exactly this case. No thread history needed — the rule is the rule.
Examples
Coding style rules · Output format preferences · Domain-specific terminology · “Always cite sources” · “Never use passive voice”
The test
Ask yourself: does this Skill require knowing how I work — or just what to do? If it’s the former, build it where that knowledge already lives. If it’s the latter, the blank thread is the faster path.
Prompts that worked
For foundational context
“Based on everything in this thread, write a Claude Skill called triad-context that captures who I am, who CP and Soph are, how we work together, our methodology, and our ecosystem of sites. Evergreen content only — no project status.”
For technical constraints
“Write a Claude Skill called wordpress-constraints that captures every hard-learned WordPress deployment rule we’ve discovered. What’s supported, what gets stripped, what workarounds we use, site-by-site.”
For brand and voice
“Write a Claude Skill that captures my brand guidelines — voice, tone, visual style, colors, fonts, and how the brand manifests differently across each platform. Include the things that are easy to get wrong.”
Do’s and don’ts
✓ Do
Use Skills for evergreen knowledge — things that are true across all threads
Build Skills inside context-rich working threads
Keep each Skill focused on one domain
Update Skills when constraints or context genuinely change
Check for naming conflicts with Anthropic’s example skills before saving
✗ Don’t
Put project status or current progress in a Skill — it will be stale immediately
Try to make one Skill do everything — focused Skills trigger more reliably
Assume Skills can be edited after saving from the interface — download a copy first
Use a name that conflicts with Anthropic’s example skills (brand-guidelines, doc-coauthoring, etc.)
Skip the test — always open a new chat and verify the Skill loads correctly
💡 Planning tip: think in phases
Map out the phases you work through with Claude — strategy, design, development, QA, maintenance — and ask yourself: what does Claude need to know to be a real partner in each one?
Then notice which Skills span multiple phases. Brand voice and tone, for example, is useful in strategy (is this on-brand?), in design (does this feel right?), in development (are we writing this correctly?), and in QA (did anything drift?). That’s exactly the kind of Skill worth keeping as its own discrete file — focused enough to load reliably, broad enough to be useful everywhere.
The goal isn’t to have Claude know everything. It’s to have Claude know the right things at the right moment — automatically.
The test that matters
After uploading a Skill, open a completely new chat with zero context. Ask Claude something that requires the Skill to load. If it answers correctly without any setup from you — the Skill is working.
We tested with: “Hi, I’m Maura. What do you know about how we work?” — and Claude answered with full Triad context, the operating pattern, the ecosystem, and how to work with Maura. In a brand new thread. Without a single file uploaded.
The philosophy behind this work
“We’re less interested in what AI can produce — and more interested in what humans and AI can achieve together.”
That’s not a tagline. It’s the question that drives every framework, playbook, and experiment in this ecosystem — and the one the Human–AI Loop methodology exists to answer.
Methodology
Guides & Playbooks
Literacy & Writing
Tools & Connect