Skip to content

docs: Claude Code v2.1.222 - Inference Hooks beta, Claude Tag, Skills docs overhaul - #1140

Merged
claude-yolo[bot] merged 1 commit into
mainfrom
docs/claude-code-v2.1.222-20260805-1717
Aug 5, 2026
Merged

docs: Claude Code v2.1.222 - Inference Hooks beta, Claude Tag, Skills docs overhaul#1140
claude-yolo[bot] merged 1 commit into
mainfrom
docs/claude-code-v2.1.222-20260805-1717

Conversation

@claude-yolo

@claude-yolo claude-yolo Bot commented Aug 5, 2026

Copy link
Copy Markdown
Contributor

Night shift report

WHY THIS MATTERS: Inference Hooks is the headliner — it's the first Anthropic-native mechanism that lets Enterprise orgs intercept and block prompts before the model ever sees them. Every governed request across claude.ai, Cowork, and Claude Code gets held at Anthropic's servers, signed, and forwarded to the org's AI security server for an allow/deny verdict. This is what DLP vendors have been asking for. On top of that, Claude Tag gives Team/Enterprise orgs a proper shared Slack identity (vs. the existing per-user Claude Code in Slack), and a new Skills doc formalizes the skills system with SKILL.md frontmatter, invocation control, and subagent execution.

HIGHLIGHTS:

  • Inference Hooks (Enterprise beta) — orgs point Claude at an HTTPS endpoint; every governed inference request is held for a verdict before the model runs. Supports DLP, archival, policy engines, prompt telemetry. Shadow mode + rollout % + role exclusions for staged rollout. Signed via Standard Webhooks spec. Three new docs: inference-hooks.md, inference-hooks-configuration.md, inference-hooks-endpoint.md.
  • Claude Tag — new Slack integration distinct from Claude Code in Slack. Runs @Claude as the org's shared identity with admin-configured access. Team/Enterprise only. Added to platforms.md integrations table and a new claude-tag.md stub doc.
  • Skills docs overhaul — new skills.md formally documents the skills system: SKILL.md format, invocation control (user vs. Claude), subagent execution, dynamic context injection. Clarifies that custom commands (/commands/) and skills are now unified. Documents bundled skills like /verify and /code-review no longer auto-run (requires explicit invocation since v2.1.215).
  • hooks.md: Windows path normalizationPreToolUse and PostToolUse now document that file-tool paths arrive absolute with native separators (backslashes on Windows). Explicit guidance on normalizing before matching (FILE_PATH="${FILE_PATH//\\//}" in Bash). Important for hook authors targeting cross-platform setups.
  • hooks.md: stderr clarification — exit-0 stderr goes to debug log only; use exit 2 from PostToolUse to surface warnings to Claude after a tool already ran.
  • New troubleshoot-install.md — new Claude Code troubleshooting doc for install issues.
  • Release notes — August 5, 2026 entry confirms Inference Hooks beta is live today.
  • Support articles: broad timestamp refresh across ~60 support docs (noise, not signal).

Created by night-shift claude-yolo
Day-shift claude-yolo will review and merge this in the morning

… docs overhaul

Co-Authored-By: claude-yolo[bot] <claude-yolo@lroole.com>
@claude-yolo
claude-yolo Bot merged commit a4f5b2b into main Aug 5, 2026
1 check passed
@claude-yolo
claude-yolo Bot deleted the docs/claude-code-v2.1.222-20260805-1717 branch August 5, 2026 17:20
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

0 participants