Skill: check-anthropic

Trigger: /check-anthropic or "check for new Anthropic features" or "what's new in Claude Code"

Description: Fetches the Claude Code GitHub releases page and official docs to find new hooks, CLAUDE.md features, and skills API changes. Cross-references against what this project currently uses and rates each gap for relevance. Run every few weeks or after a major Claude Code release.

Allowed Tools: WebFetch, Read, Glob

Steps

1. Fetch what's new

Fetch these three pages in parallel:

• https://github.com/anthropics/claude-code/releases — recent releases, changelogs
• https://code.claude.com/docs/en/hooks — full current hook event list
• https://code.claude.com/docs/en/memory — CLAUDE.md + @imports + rules/ features

Read each page and extract:

• Any new hook event names not previously known
• Any new CLAUDE.md syntax or frontmatter fields
• Any new skills API capabilities
• Any new settings.json options

2. Read what this project currently uses

• Read .claude/settings.json (project-level hooks)
• Read ~/.claude/settings.json (global hooks, if exists)
• Glob .claude/skills/*/SKILL.md — list all skills, count them
• Read CLAUDE.md — note any @imports and rules/ files in use

Extract: all hook event names currently wired.

3. Cross-reference against known full hook list

The full official hook list as of last verification (25 events): SessionStart, UserPromptSubmit, PreToolUse, PermissionRequest, PostToolUse, PostToolUseFailure, Notification, SubagentStart, SubagentStop, TaskCreated, TaskCompleted, Stop, StopFailure, TeammateIdle, InstructionsLoaded, ConfigChange, CwdChanged, FileChanged, WorktreeCreate, WorktreeRemove, PreCompact, PostCompact, Elicitation, ElicitationResult, SessionEnd

Compare: hooks in use vs full list → identify unused hooks. Also check: any hooks on the fetched docs page that are NOT in the list above → those are new.

4. Rate each gap

For every unused hook (and any new features found), rate relevance:

Guidance for rating hooks:

• PostCompact — High: re-inject memory context after compaction fires
• SessionEnd — High: complement to Stop, fires on clean exit
• FileChanged — Medium: more granular drift detection than PostToolUse Edit|Write
• StopFailure — Medium: capture state when Claude errors out
• SubagentStop — Low unless this project uses subagents
• TaskCompleted — Medium: hook into task completion for auto-logging
• TeammateIdle — Low: multi-user setups only
• ConfigChange — Low: fires when settings change

5. Report

Output in this format:

## Claude Code Feature Check — [date]

### Releases scanned
[list any releases found with key changes]

### Hook coverage
Currently using: N / 24 hooks
[list hooks in use]

### High-priority gaps
[hook name] — [one line: what it does + why it's relevant here]

### Medium gaps
[hook name] — [one line]

### New features found (not in baseline)
[anything on the docs page not in the known list above]

### CLAUDE.md / Skills API
[any new syntax, frontmatter fields, or skills features found]

### Recommendation
[1-2 sentences: what to investigate next]

Notes

• Run every 4-6 weeks, or after any Claude Code release announcement
• Pairs well with /evolve — run check-anthropic first, then evolve to implement anything rated High
• If a new hook is rated High, create a plan with Plan [hook name integration] before implementing
• Update the known hook list in Step 3 if new official hooks are found