GitHub agent-skill-group search diagnostics
GitHub repository search can treat agent-skill-group as a broad token query: agent, skill, and group.
That is why a search that visually looks identical to the project name can still show unrelated or more established repositories first. The canonical repository is go165/agent-skill-groups; the transparent alias repositories point back to it.
What to check
| Question | Command | Expected signal |
|---|---|---|
| Can GitHub find the exact repository name? | gh search repos "agent-skill-groups in:name" --limit 20 | go165/agent-skill-groups should appear at or near the top. |
| Can users find the project from intent terms? | gh search repos "github skill-group" --limit 20 | One of the canonical or alias repositories should appear. |
| Which broad matches are outranking the project? | agent-skill-groups discovery --json | Read rankPressureQueries, projectFamilyRank, canonicalRank, aliasRanks, nonTargetTopResults, matchSignals, and searchIntent. |
| Which query should I copy when broad search is noisy? | agent-skill-groups in:name | The report exposes this as the stable query for exact repository-name discovery. |
Run the diagnostic report
agent-skill-groups discovery --json --required-rank 10
agent-skill-groups discovery-report --output-md DISCOVERY_REPORT.md --output-html discovery-report.html --jsonThe generated report includes backend, project-family rank, canonical rank, alias ranks, rank pressure summary, top results, ranking diagnostics, and the top non-target repositories with match signals. A transparent alias repository can help users navigate to the canonical project, but it should not be counted as the main repository's own rank.
When the same search term shows other projects first
For a plain query such as agent-skill-groups, inspect searchIntent.kind, searchIntent.stableQuery, rankingDiagnostics.nonTargetCountAheadOfProjectFamily, rankingDiagnostics.nonTargetCountAheadOfCanonical, rankingDiagnostics.aliasCountAheadOfCanonical, and rankingDiagnostics.highestStarNonTargetAhead. Those fields separate unrelated repositories from this project's transparent alias repositories and show whether popularity signals are likely dominating an exact-looking name match.
If searchIntent.kind is broad-relevance or broad-observation, use the stable query shown by the report. For the canonical repository, that stable query is agent-skill-groups in:name.
The report also publishes sortModeComparisons for agent-skill-groups. Check this field when the same search text appears to behave differently in GitHub's default repository search, exact-name search, updated sorting, and star sorting.
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"Why this matters
The project is not trying to hide weak discovery behind a hand-picked query. It publishes the current state, including broad queries where ranking is still poor. That makes the search problem measurable: exact-name checks, intent-query checks, broad-query observations, Pages availability, and ordinary web indexing can all be tracked separately.