Conversation
…e/skills Skills authored for Claude Code live in ~/.claude/skills and project .claude/skills, which step never scanned, so users had to point the model at them by hand. Discover both alongside .agents/skills: the user directory always, project directories from cwd up to the git root once the project is trusted. A .claude/skills entry symlinked into ~/.agents/skills is loaded once. Also auto-discover ~/.stepcode/skills as user skills next to ~/.stepcode/agent/skills. When cwd is $HOME that directory is the user's own, so it is not read again as project skills and does not trigger the project trust prompt. Fixes #205
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Fixes #205
Problem
Step never scanned Claude Code's skill directories, so skills users already keep in
~/.claude/skills(often symlinked from~/.agents/skills) or in a project's.claude/skillswere invisible until someone pointed the model at them by hand.~/.stepcode/skillswas not auto-discovered either; only~/.stepcode/agent/skillswas.Changes
Skill discovery (
core/package-manager.ts)~/.claude/skillsand~/.stepcode/skillsare now loaded alongside~/.stepcode/agent/skillsand~/.agents/skills, without needing project trust..claude/skillsis scanned from cwd up to the git root, the same way.agents/skillsis, once the project is trusted..claudeand.agentsfollow the same scan rules.~/.stepcode/skillsfollows the same rules as~/.stepcode/agent/skills.$HOME,~/.stepcode/skillsand~/.claude/skillsload once, as user skills, and are not read again as project skills.~/.claude/skillssymlinks into~/.agents/skillsloads once.~/.stepcode/agent/skills→~/.stepcode/skills→~/.agents/skills→~/.claude/skills. When two skills share a name, the first one loaded wins.Project trust (
core/trust-manager.ts).claude/skillstriggers the trust prompt, just as.agents/skillsdoes.~/.claude/skillsand~/.stepcode/skillscount as the user's own resources, so starting in$HOMEdoes not ask for trust.Docs: skill locations and the trust notes are updated in
docs/skills.md,sdk.md,security.md,settings.md,usage.md,extensions.mdandREADME.md.Behavior change
When cwd is
$HOME,~/.stepcode/skillsused to be read as project skills and needed project trust. It is now a user skill directory and always loads. The existing testshould resolve symlinked user and project resources oncenow expects the skill's scope to beuser.Tests
.claude/skills, the symlink dedupe,~/.<configDir>/skillsdiscovery with a single load when cwd is$HOME, and trust detection.packages/coding-agent: 3736 passed, 51 skipped.apps/cli: 631 passed.pnpm run checkpasses.