This release adds 3 notable features for engineering teams evaluating rollout.
✓ No known CVEs patched in this version
Topics
+2 more
Affected surfaces
Summary
AI summaryUpdates Dependency updates, What's new, and Tool annotations across a mixed release.
Full changelog
What's new
MCP Prompts (slash commands)
Skills now appear as slash commands in MCP clients (Claude Desktop, Cursor, etc.). No tool calls needed — select a skill from the slash command menu to inject it directly into context.
Live skill updates
File watcher (chokidar) watches all skill directories. Add or remove a SKILL.md and connected clients receive a notifications/prompts/list_changed notification automatically. Disable with MCP_NO_WATCH=1.
Skill search/filter
list_skills now accepts an optional query parameter — case-insensitive substring match on name and description. Useful with large skill libraries.
Tool annotations
readOnlyHint: trueonlist_skills,read_skill,debug_info,manage_search_pathsdestructiveHint: trueoninstall_skill
Dependency updates
@modelcontextprotocol/sdk:^1.0.1→^1.29.0typescript:^5.4.5→^5.8.0@types/node:^20→^22- Added
chokidar ^5.0.0
Bug fix
Server version was hardcoded as "1.0.0" — now reads from package.json at runtime.
Upgrade
npm install -g agent-skill-loader
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Share this release
About back1ply/agent-skill-loader
Dynamically load Claude Code skills into AI agents without copying files. Discover, read, and install skills on demand.
Related context
Beta — feedback welcome: [email protected]