GitHub search ranking for agent-skill-groups
Exact repository-name discovery and broad GitHub relevance ranking are different checks.
The canonical repository is go165/agent-skill-groups. A plain GitHub search for agent-skill-groups can still rank older repositories first because GitHub tokenizes the query and weighs stars, freshness, descriptions, topics, and repository authority.
Recommended checks
agent-skill-groups discovery --json
agent-skill-groups discovery --json --fail-on-missing
agent-skill-groups discovery --json --required-rank 10
gh search repos "agent-skill-groups in:name" --limit 20
gh search repos "agent-skill-groups" --limit 20
gh search repos "github skill-group" --limit 20
gh search repos "github agent-skill-group" --limit 20agent-skill-groups in:namechecks exact repository-name discoverability.agent-skill-groupsis a broad relevance query.github skill-groupandgithub agent-skill-grouptest alias discovery.
Repeatable discovery report
The installed CLI records rankings for the canonical repository, transparent alias repositories, and Pages availability from any OS:
agent-skill-groups discovery --json
agent-skill-groups discovery --json --required-rank 10--fail-on-missing gates exact-name and intent-specific queries. --required-rank 10 fails when a required target appears below the top 10. Broad observation queries such as skill-group, skill groups, skill manager, and skills manager stay in the report but are optional by default because GitHub ranks older high-star generic repositories for those naked terms.
Each query records a backend value. The CLI prefers gh search repos because that path is closer to what users see in GitHub's repository search UI; if the GitHub CLI is unavailable or fails, it falls back to the GitHub REST Search API. Those two backends can rank the same query differently.
The report separates projectFamilyRank for the best project-owned result from canonicalRank for go165/agent-skill-groups and aliasRanks for transparent navigation repositories. This prevents an alias result from hiding the main repository's actual rank while still showing the best project entry.
Each query also includes rankingDiagnostics. Use nonTargetCountAheadOfProjectFamily to count unrelated repositories ahead of the best project-owned result, nonTargetCountAheadOfCanonical to count unrelated repositories ahead of the canonical result, aliasCountAheadOfCanonical to count this project's transparent aliases ahead of it, and highestStarNonTargetAhead to identify the strongest unrelated result above it.
The default discovery run also compares the same visible query across GitHub sort modes. This separates exact-name discovery from default relevance, freshness, and star-weighted ranking. A new project can be first for agent-skill-groups in:name and agent-skill-groups sort:updated, while agent-skill-groups sort:stars still places many older repositories ahead of it.
agent-skill-groups discovery --json --query "agent-skill-groups"
agent-skill-groups discovery --json --query "agent-skill-groups" --query "agent-skill-groups in:name" --query "agent-skill-groups sort:updated" --query "agent-skill-groups sort:stars"
gh search repos "agent-skill-groups in:name" --limit 20The PowerShell compatibility script remains available for Windows environments and optional ordinary Google sampling:
powershell -ExecutionPolicy Bypass -File .\scripts\check-discovery.ps1 -Limit 20Current visibility strategy
The project uses the exact repository name, accurate topics, GitHub Pages, structured data, a sitemap, transparent alias repositories, release assets, and external directory PRs. These signals improve discoverability over time, but GitHub and Google ranking cannot be forced immediately from the repository.