Claude Report — 2026-05-08
- Claude Code 2.1.133 ships effort-level injection in hooks (`$CLAUDE_EFFORT`), `worktree.baseRef` control, and custom Linux sandbox binary paths.
- Anthropic Python SDK v0.100.0 (May 6) adds complete Managed Agents SDK surface: multiagent sessions, outcomes, webhooks, vault validation.
- Plugin marketplace policy scanner upgraded to blocking: hook-scope overreach, undisclosed telemetry, and description mismatches now auto-reject marketplace submissions.
- Eight new plugins (ClickHouse, qdrant-skills, zilliz, dash0, twilio-developer-kit, and more) enter the marketplace; three removed.
🚀 Anthropic Official
Claude Code 2.1.128–2.1.131 and 2.1.133 — May 5–7
Adds worktree.baseRef: fresh | head for --worktree, EnterWorktree, and agent-isolation worktree base selection; sandbox.bwrapPath / sandbox.socatPath managed settings for custom Linux/WSL bubblewrap and socat binaries; parentSettingsBehavior admin key for SDK managedSettings policy merge; and effort-level exposure in hooks via effort.level JSON input and $CLAUDE_EFFORT env var (2.1.133). Earlier versions ship --plugin-url for URL-based session plugin loading, .zip archive support for plugins, /color for random session colors, and fixes for VS Code Windows extension activation and Mantle endpoint auth headers (2.1.128–2.1.131).
anthropic-sdk-python v0.99.0–v0.100.0 — May 5–6
v0.100.0 (May 6) adds full Managed Agents SDK surface — multiagent sessions, outcomes, webhooks, and vault credential validation — providing Python bindings for the API features that reached public beta the same day. v0.99.0 (May 5) adds workspace targeting for OIDC federation token exchange on the client.
🔌 Claude Code Plugins
None of the top 10 updated in window.
Unchanged in window: frontend-design, superpowers, context7, code-review, code-simplifier, github, skill-creator, playwright, feature-dev, claude-md-management.
🛠️ Skills
None of the top 5 updated in window.
Unchanged in window: superpowers:brainstorming, superpowers:tdd, superpowers:systematic-debugging, superpowers:writing-plans, superpowers:executing-plans.
🤖 Agents & Subagents
None of the top 5 updated in window.
Unchanged in window: superpowers:subagent-driven-development, claude-router:orchestrate, ralph-loop, claude-router:fast-executor, claude-mem:do.
🔗 MCPs & Integrations
None of the top 5 updated in window.
Unchanged in window: github-mcp-server, modelcontextprotocol/servers, filesystem MCP, claude-mem, context7.
💡 Community — Workflows & Ideas
claude-plugins-official — Policy scanner upgraded to blocking (PR #1771, May 7)
Raises the marketplace admission bar: adds mandatory hook-scope analysis (has_broad_scope_hooks), undisclosed-telemetry detection (has_undisclosed_telemetry), and description-behavior mismatch checks — any violation now fails the scan and blocks the PR. All 117 existing external plugins will be re-scanned under the new rules; authors with broad-scope hooks or default-on telemetry without explicit opt-out disclosure must update before resubmission.
Marketplace batch: 8 added, 3 removed — May 6–7
Adds ClickHouse, pigment, qdrant-skills, zilliz, dash0, outputai (vector-DB and observability cluster), plus twilio-developer-kit and cwc-makers (M5Stack Cardputer hardware onboarding with curl+tar fallback when git is absent). Removes flint, optibot, and versori-skills.
📰 Quick Mentions
- Releasebot: Claude Code May 2026 — Continuous tracking of Claude Code releases with first-seen timestamps; useful for catching patch versions between report windows.
- quemsah/awesome-claude-plugins — n8n-powered automated tracker of Claude Code plugin adoption across GitHub; updated May 7, 15,861 repos indexed.