feat(skills): five engineering & design sources join the featured list (v0.330.0) - #601
Merged
Conversation
…t (v0.330.0) The console's "Install from GitHub" dialog shipped seven featured sources, six of them marketing. Agents doing engineering work had to know a repo name to type. Adds five, each verified to resolve through `browseRepo` first: addyosmani/agent-skills (24), google/skills (88), emilkowalski/skills (9), Nutlope/hallmark (anti-slop design) and tt-a1i/archify (diagrams). A preset is only a pointer — nothing is vendored, the repo's SKILL.md folders are listed at click time and an owner/admin picks what to install — so this is reach, not new trust surface. That verification step earned its keep: msitarzewski/agency-agents (140k stars, trending) has zero SKILL.md folders and would have shipped a chip that browses to an empty list. Also corrects the existing counts. They had drifted (anthropics 157k→167k stars, mattpocock 151k→211k), and the raw SKILL.md count overstates what a user gets: browseRepo dedupes by folder name, so alirezarezvani/claude-skills is 439 installable, not 798. Every count is now the deduped figure the dialog shows. scripts/skill-presets-test.cjs pins the two mechanical failure modes — a repo that isn't the plain owner/repo form, and a duplicate entry — offline, so the suite stays hermetic. Live resolution stays a manual check documented in the file's header, since it needs GitHub's unauthenticated 60/hr budget. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DiKurWa52Lm3CJrbTdCa2W
vikasprogrammer
added a commit
that referenced
this pull request
Aug 11, 2026
…t (v0.330.0) (#601) The console's "Install from GitHub" dialog shipped seven featured sources, six of them marketing. Agents doing engineering work had to know a repo name to type. Adds five, each verified to resolve through `browseRepo` first: addyosmani/agent-skills (24), google/skills (88), emilkowalski/skills (9), Nutlope/hallmark (anti-slop design) and tt-a1i/archify (diagrams). A preset is only a pointer — nothing is vendored, the repo's SKILL.md folders are listed at click time and an owner/admin picks what to install — so this is reach, not new trust surface. That verification step earned its keep: msitarzewski/agency-agents (140k stars, trending) has zero SKILL.md folders and would have shipped a chip that browses to an empty list. Also corrects the existing counts. They had drifted (anthropics 157k→167k stars, mattpocock 151k→211k), and the raw SKILL.md count overstates what a user gets: browseRepo dedupes by folder name, so alirezarezvani/claude-skills is 439 installable, not 798. Every count is now the deduped figure the dialog shows. scripts/skill-presets-test.cjs pins the two mechanical failure modes — a repo that isn't the plain owner/repo form, and a duplicate entry — offline, so the suite stays hermetic. Live resolution stays a manual check documented in the file's header, since it needs GitHub's unauthenticated 60/hr budget. Claude-Session: https://claude.ai/code/session_01DiKurWa52Lm3CJrbTdCa2W Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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.
What
The console's Install from GitHub dialog shipped seven featured sources, six of them marketing. This adds five engineering/design libraries, and corrects the counts on the existing seven.
addyosmani/agent-skillsgoogle/skillsemilkowalski/skillsNutlope/hallmarktt-a1i/archifyWhy this is reach, not new trust surface
A preset is only a pointer. Nothing is vendored into the repo — clicking a chip calls
browseRepo, which lists that repo'sSKILL.mdfolders live, and an owner/admin picks what to install. An installed skill is then governed identically to a hand-authored one: files in the tenant's library, materialised at launch, every effect still through the PreToolUse gate hook.Verification
Each candidate was resolved through
browseRepobefore being added. That step earned its keep:msitarzewski/agency-agents(140k stars, currently trending) contains zeroSKILL.mdfolders — it's a repo of agent personas — and would have shipped a chip that browses to an empty list. Rejected.All 12 presets resolve:
The count fix
The hand-written counts had drifted (
anthropics/skills⭐157k→167k,mattpocock/skills⭐151k→211k). More usefully, the rawSKILL.mdcount overstates what a user gets:browseRepodedupes skills by folder name (first wins), soalirezarezvani/claude-skillsis 439 installable, not the 798 files in the tree. Every count is now the deduped figure the dialog itself shows.Test
scripts/skill-presets-test.cjs(wired intonpm run test:governance, 51 checks) pins the two mechanical failure modes — arepothat isn't the plainowner/repoform, and a duplicate entry — offline, so the suite stays hermetic. Falsified by injecting a URL-form repo: fails with the right message, exit 1. Live resolution stays a manual check documented in the file header, since it needs GitHub's unauthenticated 60/hr budget.npm run typecheck✅ ·cd web && npm run build✅ ·npm run test:governance→ 14 passed, 0 failed ✅🤖 Generated with Claude Code
https://claude.ai/code/session_01DiKurWa52Lm3CJrbTdCa2W