issue-hygiene-reminder

A command hook for Claude Code and Codex. Before a GitHub issue write it adds one paragraph to the model's context: check the write was authorized, the body leads with current state and one next action, evidence lives in the repo or the issue, at most one open fork, no decision checklists, no agent names, labels per the repo's meanings, and a close is one or two sentences. It names the issue-hygiene skill in personal-skills for the full rules. It never blocks.

It fires on gh issue create|edit|close|reopen|comment|delete|transfer|lock|unlock|pin|unpin|develop, on gh api ... issues ... with a write flag, and on a GitHub MCP tool whose name contains issue and a write verb (create_issue, update_issue, add_issue_comment, ...). Reads, lists, and searches stay silent.

Why a hook and not only a skill: a skill loads when the model decides to load it; the hook fires at the moment that matters, whatever the model remembered.

Codex skips a plugin's hooks until you review and trust them (/hooks in Codex). Claude Code runs them once the plugin is enabled.

Test: sh plugins/issue-hygiene-reminder/tests/remind.test.sh (needs jq).