You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Draft implementation complete and independently validated. Kept draft because repository administration must still require pull requests and the Catalog Manifest status check on main.
Summary
add a typed 11-source recommended skill catalog, separate from marketplace registry state
add allagents skill search <query> --catalog recommended with a hard repository/subpath boundary and no global fallback
preserve exact catalog source, root, ref, selectors, and provenance through project-scoped installs and updates
distinguish selectable marketplaces/subtrees from optional, search-only, and external-lifecycle sources
fix clean-clone LFS filter handling and add read-only catalog health validation
document the complete implementation plan in .claude/plans/recommended-skill-catalog.md
Sources
Includes the requested gstack, Paperclip, Matt Pocock, Composio, Hermes core, and Hermes optional sources plus broadly reusable candidates verified from ai-research-wiki: Anthropic Skills, Addy Osmani Agent Skills, Superpowers, Context Engineering Skills, and Anthropic Knowledge Work Plugins.
Elastic Docs Skills was deliberately excluded after review because it is Elastic-product-specific rather than broadly reusable.
Verification
130 focused tests passed; 0 failed
final affected subset: 79 passed; 0 failed
typecheck and production build passed
catalog:validate: all 11 sources healthy
unknown catalog and --catalog + --owner: exit 2
isolated clean-cache project installs exercised all installable source classes
Paperclip references preserved; Matt manifest boundary preserved; Hermes core/optional coexist without provenance collapse; sibling assets preserved
gstack and Composio non-installable selections produced no workspace/cache mutation
Implementation and independent validation are now pushed. The review fixed mixed-policy install grouping, offline user-sync provenance, and catalog pagination starvation. Elastic was removed from the catalog as product-specific. Remaining blocker is repository administration: require PRs and the Catalog Manifest status check on main before merge.
CI repair and independent full-diff review pushed. Root causes fixed: Biome non-null assertion, Bun suite-wide simple-git mock contamination, and a 5s build hook timeout. Independent review then found and fixed catalog policy-metadata spoofing, all-sections-failed messaging, unsupported simple-git options, subprocess isolation bounds, import-time test builds, unbounded catalog health requests, and stale docs/plan claims.
Post-review CI-equivalent verification on Bun 1.3.12: lint passed; 1,500 tests passed / 5 skipped / 0 failed; typecheck passed; build passed; all 11 catalog sources healthy.
Follow-up CI fix pushed: the LFS clone test now derives its expected pointer from the repository's committed blob after Git clean filters run. This removes environment dependence on whether git-lfs is installed while still proving clone skips smudge and does not materialize payload bytes. Bun 1.3.12 full suite: 1,500 passed / 5 skipped / 0 failed; lint and typecheck passed.
Second runner-specific LFS fixture issue fixed: CI had git-lfs installed, so the fixture's pre-push hook rejected an intentionally incomplete pointer before cloneTo ran. The test now clones directly from the committed non-bare fixture, avoiding unrelated push hooks while preserving all clone/LFS assertions. Verified with Git LFS 3.7.1 explicitly enabled, full Bun 1.3.12 suite 1,500/0 failures, lint and typecheck.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Status
Draft implementation complete and independently validated. Kept draft because repository administration must still require pull requests and the
Catalog Manifeststatus check onmain.Summary
allagents skill search <query> --catalog recommendedwith a hard repository/subpath boundary and no global fallback.claude/plans/recommended-skill-catalog.mdSources
Includes the requested gstack, Paperclip, Matt Pocock, Composio, Hermes core, and Hermes optional sources plus broadly reusable candidates verified from
ai-research-wiki: Anthropic Skills, Addy Osmani Agent Skills, Superpowers, Context Engineering Skills, and Anthropic Knowledge Work Plugins.Elastic Docs Skills was deliberately excluded after review because it is Elastic-product-specific rather than broadly reusable.
Verification
catalog:validate: all 11 sources healthy--catalog+--owner: exit 2