Skill Crossroads
Skill Crossroads is an evidence-based quality and safety grader for Claude Code artifacts, including skills, subagents, slash commands, MCP configurations, and plugins. It produces letter grades, file-and-line evidence, embeddable badges, and prioritized fixes so builders can decide whether to ship, fix, or rethink an artifact.
At a Glance
- Developers and teams building Claude Code skills and agents
- Maintainers of slash commands, MCP configurations, and Claude Code plugins
- Open-source projects publishing AI coding artifacts on GitHub
- Engineering organizations that want private-repository quality checks and shared dashboards
- +1 more
AI Tools by Skill Crossroads
(1)
Skill Crossroads
Claude Code Artifact Grader
Discussions
No discussions yet
Be the first to start a discussion about Skill Crossroads
Latest News
Show HN launch: Linting 216 public Claude Code skills; report says 69% would not reliably trigger
Published The State of Claude Code Skills, an audit of 216 public skills across 18 repositories
Added the init command to insert an always-fresh Skill Crossroads badge into a repository README
Public Skill Crossroads repository created with the evidence-cited Claude Code artifact grading project
Products & Services
Local command-line scanner for Claude Code skills and related artifacts. It runs deterministic checks, can optionally run LLM-assisted checks with the user's own Anthropic key, and can generate HTML scorecards and SVG badges.
Web product for scanning public GitHub repositories or pasted artifacts and displaying letter grades, six category scores, file-and-line evidence, prioritized fixes, and shareable always-fresh badges.
CI integration that scans artifacts on pull requests, posts scorecards and inline annotations, and can fail a build below a configured minimum grade.
Paid hosted tier for private-repository scanning, managed LLM checks, hosted scorecards and always-fresh badges, and score history.
Market Position
Skill Crossroads positions itself specifically around craftsmanship, discoverability, safety, and evidence-cited scorecards for Claude Code artifacts. Its repository contrasts the product with Anthropic's skill-creator, which measures effectiveness, while Skill Crossroads grades whether an artifact is well-built, discoverable, safe, and ready to ship; adjacent alternatives include generic code linters and manual review, which do not provide this artifact-specific rubric-and-badge workflow.
Leadership
Founders
Steve Harlow (GitHub: sgharlow)
VP Engineering focused on AI-assisted code that shows its work; GitHub identifies him as the creator/owner of the Skill Crossroads repository and lists him in Arizona, USA.
Executive Team
Steve Harlow
Creator / maintainer; VP Engineering
GitHub profile describes him as a VP Engineering working on AI-assisted code that shows its work; he owns and maintains the Skill Crossroads repository under sgharlow.
Founding Story
The project was started after its creator identified recurring real-world failures in Claude Code artifacts—especially skills that never trigger, over-broad tool grants, hardcoded secrets, and missing tests. The initial vision was an open-source linter and evidence-cited scorecard that checks what developers actually ship, with a free CLI/public-scan path and hosted paid features for private repositories and managed checks.
Business Model
Revenue Model
Freemium SaaS and open-core model: the CLI, public scans, deterministic rubric, local reports/badges, and GitHub Action are free; paid subscriptions provide private-repository scanning, managed LLM checks, hosted scorecards, score history, and organization features. Billing is handled through Stripe.
Pricing Tiers
CLI and public-repository scans, full deterministic rubric, GitHub Action and PR gating, bring-your-own-key triggering/token checks, local HTML report, and SVG badge.
Everything in Free plus private-repository scanning, managed LLM with no user key required, hosted scorecards and always-fresh badges, and score history; 14-day trial.
Everything in Pro plus organization-wide custom rules, team seats, and shared dashboard.
Target Markets
- Developers and teams building Claude Code skills and agents
- Maintainers of slash commands, MCP configurations, and Claude Code plugins
- Open-source projects publishing AI coding artifacts on GitHub
- Engineering organizations that want private-repository quality checks and shared dashboards
- Teams using CI/CD and pull-request governance for AI-assisted code
- Pre-release quality and safety review of Claude Code skills
- Checking whether descriptions reliably trigger model invocation
- Auditing subagent tool permissions and MCP configuration security
- Reviewing Claude Code plugins and their manifests, components, and hooks
- Generating shareable quality badges for README files
- Preventing quality regressions through GitHub pull-request gating
- Anthropic's public skills repository is used as a report sample and as a hosted scorecard example.
- The creator's sgharlow/claude-code-recipes repository uses a Skill Crossroads badge and is referenced as a live scorecard example.