[11.5.0] - 2026-05-21 - "Security Boundary Hardening and New Agent Skills"
Installable skill library update for Claude Code, Cursor, Codex CLI, Gemini CLI, Antigravity, and related AI coding assistants.
This release hardens plugin-safe distribution boundaries, resolves the current Dependabot ws alert in the web app lockfile, and merges PR #604 and PR #605 through the maintainer workflow.
New Skills
- subagent-orchestrator - quota-aware parallel subagent coordination for large, multi-file Antigravity tasks.
- bilig-workpaper - formula-backed WorkPaper JSON and MCP guidance for deterministic spreadsheet-style agent workflows.
Security
- plugin-safe boundary hardening - blocks critical-risk or third-party setup-sensitive community skills from the generated Codex and Claude plugin-safe distributions.
- ingest-youtube input hardening - validates single-video YouTube URLs, terminates
yt-dlpoptions with--, ignores user config, adds subprocess timeouts, and neutralizes untrusted metadata in generated markdown. - web app dependency fixes - updates
wsto the patched8.20.1range and refreshes vulnerable transitive audit entries inapps/web-app/package-lock.json.
Improvements
- generated artifact sync - refreshes compatibility data, skill indexes, plugin mirrors, docs, and visible skill counts to
1,464+. - subagent-orchestrator metadata - normalizes PR #604 metadata so the skill passes repository validation.
Who should care
- Plugin users get a safer marketplace-style distribution that excludes setup-sensitive critical workflows from safe plugin bundles.
- Vault users get safer YouTube transcript ingestion with stricter URL and markdown handling.
- Agent workflow builders get two new community skills for parallel agent coordination and formula-backed workbook automation.
Credits
- @sulavmgr456-byte for PR #604 (
subagent-orchestrator). - @gregkonush for PR #605 (
bilig-workpaper).