github-check

Show GitHub PR notification details from cache.

When to use

• When the status line shows a GitHub count and you want to see the PR list.
• When the status line shows ! for GitHub and you want to diagnose the error.
• When you want to force an immediate refresh instead of waiting for the TTL.

Steps

1. Read the cache file

CLAUDE_PLUGIN_DATA="${CLAUDE_PLUGIN_DATA:-$HOME/.claude/plugins/data/claude-code-status-claude-code-status}"
cat "$CLAUDE_PLUGIN_DATA/cache/github.json"

Parse the JSON. The items array contains the detailed PR notification list.

2. If status is ok — display the items

The items array contains objects with:

Field	Example
title	Optimize database queries (#312)
link	https://github.com/myorg/backend/pull/312
meta.repo	myorg/backend
meta.reason	review_requested
meta.updated	2026-03-24T10:10:00Z

Present as a numbered list, most recently updated first. Include the link for each PR.

3. If status is error — show error cause

errorKind	Likely cause	Recommended fix
auth	Not logged in via gh	Run gh auth login
dependency	gh CLI not installed	Install from https://cli.github.com
rate_limit	GitHub API rate limit exceeded	Wait a minute, then retry
transient	Temporary network error	Force refresh

4. Force refresh (if user requests)

CLAUDE_PLUGIN_DATA="${CLAUDE_PLUGIN_DATA:-$HOME/.claude/plugins/data/claude-code-status-claude-code-status}"
rm -f "$CLAUDE_PLUGIN_DATA/locks/github.lock"
node "$CLAUDE_PLUGIN_DATA/runtime/dist/collect.js" --service github --force 2>&1
cat "$CLAUDE_PLUGIN_DATA/cache/github.json"

Then display the updated items.

Example output

GitHub unread PR notifications: 4   (last updated: 2026-03-24T10:15:00Z)

 1. myorg/backend  Optimize database queries (#312)
    Reason: review_requested  Updated: 2026-03-24T10:10:00Z
    https://github.com/myorg/backend/pull/312

 2. myorg/frontend  Add dark mode toggle (#298)
    Reason: mention  Updated: 2026-03-24T09:45:00Z
    https://github.com/myorg/frontend/pull/298

Notes

• Data is pre-collected with details — no additional API calls needed.
• The count uses thread-level deduplication.
• Cache TTL: 1 minute.