feat: improve product skill score (41% → 90%)

[yogesh-tessl]

Hey @kmcginnes 👋

ran your skills through tessl skill review at work and found some targeted improvements for the product skill. Here's the before/after:

Skill	Before	After	Change
product	41%	90%	+49%

I focused on the product skill since it had the most room to grow. It's the "front door" overview of Graph Explorer and was scoring lowest alongside schema.

Changes made to product skill

• Added explicit "Use when..." clause to the description so agents know when to select this skill
• Removed unknown tools frontmatter key that was triggering a validation warning
• Added a concrete quickstart section with copy-paste Docker commands (from your existing docs)
• Added a database/query language compatibility table with decision guidance (Gremlin vs openCypher vs SPARQL)
• Added cross-references to deeper docs (docs/features/connections.md, docs/architecture.md, etc.) for progressive disclosure
• Tightened the architecture section with monorepo structure, key libraries, and the proxy rationale
• Kept the diff small (64 lines) - all additive, no structural rewrites

---

also stress-tested your git skill against a few real-world task evals and it held up really well on Vitest hook testing with renderHookWithState and backward compatibility checks for persisted data. Kudos for that.

quick honest disclosure. I work at https://github.com/tesslio where we build tooling around skills like these. Not a pitch, just saw room for improvement and wanted to contribute.

If you want to self-improve your skills, or define your own scenarios to pressure test, just ask your agent (Claude Code, Codex, etc.) to evaluate and optimize your skill with Tessl. Ping me @yogesh-tessl, if you hit any snags.

[kmcginnes]

Thanks for the contribution @yogesh-tessl! I appreciate you taking the time to analyze the skill and propose improvements.

I like several of the changes here: the frontmatter cleanup, the tighter summary sentence, the feature list with doc cross-references, and the architecture rewrite. However, I'm requesting changes on the quick start section and the database table (see inline comments), and I'd like to keep the "Core Purpose" section.

Also worth noting: these skill files will likely be in flux as I adopt a more structured approach to AI steering, so don't be surprised if things shift around soon.

[kmcginnes]

The "Core Purpose" section describes the product's use case, which the agent can't derive from code alone. It helps the agent understand what problem Graph Explorer solves, not just what it is. I'd prefer to keep it since the vagueness is deliberate — it stays accurate as the app evolves.

[kmcginnes]

This "Quickstart" section duplicates the user guide and isn't something the agent would typically need to perform. If it ever does, it can discover the quick start instructions from the existing docs via progressive disclosure. I'd drop this section to keep the skill focused on product context rather than setup steps.

[kmcginnes]

I like the intent here, but I worry this table is too prescriptive for an agent facing doc. These are providers we've tested, not an exhaustive set of what's supported. There are other Gremlin and SPARQL compatible databases that would work fine but aren't listed. Could we rename the column to "Example Databases" to frame these as known-good examples rather than a definitive list? Otherwise the agent might reject valid configurations just because they're not in the table.