Claude Report — 2026-06-16

🚀 Anthropic Official

Claude Code v2.1.178

Released June 15. Introduces Tool(param:value) syntax so permission rules can now match on specific tool input parameters with * wildcard support (e.g., Agent(model:opus) blocks Opus subagents, Bash(cmd:rm) denies rm invocations); nearest-.claude/ directory wins on skill/agent/workflow name collisions and project-scope workflow saves target the closest .claude/workflows/; auto mode now routes each subagent spawn through the safety classifier before launch; /doctor gets a consistent flat tree layout with clearer status icons; fixes include an fd-exhaustion crash from stale WebSocket/OAuth environment variables, silent Chrome OAuth connect failure on account mismatch, and vim-mode undo (now steps one command at a time instead of replaying the full buffer).

ClaudeForFoundationModels v0.1.2

Released June 15. Adds partial-snapshot delivery during streaming responses — Apple-platform apps on iOS 27, macOS 27, iPadOS 27, and visionOS 27 now receive incremental Claude output as it generates rather than buffering to completion; previous versions delivered only the final complete result.

🔌 Claude Code Plugins

boltz — new addition

Adds protein structure prediction, molecule screening, and binder design via the Boltz-2 compute API — the first biomolecular-design plugin in the official catalog, callable directly from Claude Code without leaving the terminal.

lovable — new addition

Integrates Lovable's official OAuth 2.1 remote MCP server; exposes focused commands for building, iterating, deploying, and managing Lovable apps from within Claude Code, with safeguards against accidental credit spend or premature publication.

Unchanged in window (automated subtree pointer bump — no verifiable feature change in the window): superpowers, forge-skills, firecrawl, rc, sap-fiori-mcp-server, teamcity-cli, revenuecat, sentry, nimble, hyperframes

🛠️ Skills

claude-code-system-prompts — v2.1.178 update

Updated June 16 to track v2.1.178; records a −20,964 token reduction in Claude Code's aggregate system-prompt token count — the largest single-version trim in the project's 211-version history — pointing to significant internal prompt restructuring or consolidation in v2.1.178.

Unchanged in window: anthropics/knowledge-work-plugins, glebis/claude-skills, anthropics/skills, ComposioHQ/awesome-claude-skills

🤖 Agents & Subagents

claude-code-action v1.0.149

Released June 15. Fixes shell-quote collapsing unquoted Bash(X:\*) permission rules to bare Bash (breaking all glob-based tool restrictions in CI); separately aligns the MCP allowed-tools parser with the SDK option parser so MCP server tool lists no longer diverge from settings.

claude-agent-sdk-python v0.2.102

Released June 15. Bundles Claude CLI v2.1.178; no additional API-level changes — consumers get all auto-mode classifier and subagent crash fixes from that CLI release automatically on upgrade.

Unchanged in window: nicobailon/pi-subagents, Chachamaru127/claude-code-harness

🔗 MCPs & Integrations

None of the top 5 had verifiable updates in the window.

Unchanged in window: modelcontextprotocol/servers, github/github-mcp-server, modelcontextprotocol/typescript-sdk, awslabs/mcp, portainer/portainer-mcp

💡 Community — Workflows & Ideas

No relevant community posts with verified publication dates in the window.

📰 Quick Mentions

Relevant links that don't warrant a section of their own:

• Claude Code v2.1.177 — June 13, changelog-only commit; no feature or fix content
• Configure Permissions — Claude Code Docs — official docs page updated to reflect the new Tool(param:value) syntax from v2.1.178; canonical reference for the deny→ask→allow evaluation order and parameter-matching examples
• claude-plugins-official PRs merged June 13-16 — 104 closed PRs in window; majority are automated subtree bumps from the CI bot, with boltz (#2855) and lovable (#2852) being the only human-authored additions